INH Capital (@INHCapital) header
INH Capital (@INHCapital) profile picture

Principal investing & research for proprietary capital across semiconductors and the AI infrastructure. Check out Faustus: https://teapawt.org/t.co/lTqclTTmat

Joined September 2026
Photo by INH Capital (@INHCapital)Photo by INH Capital (@INHCapital)Photo by INH Capital (@INHCapital)Photo by INH Capital (@INHCapital)Photo by INH Capital (@INHCapital)Photo by INH Capital (@INHCapital)Photo by INH Capital (@INHCapital)Photo by INH Capital (@INHCapital)Photo by INH Capital (@INHCapital)Photo by INH Capital (@INHCapital)
๐—ก๐—ฉ๐—œ๐——๐—œ๐—” ๐—ค๐Ÿฏ ๐—˜๐—ฎ๐—ฟ๐—ป๐—ถ๐—ป๐—ด๐˜€ ๐—ฃ๐—ฟ๐—ฒ๐˜ƒ๐—ถ๐—ฒ๐˜„ | ๐—œ๐—ก๐—› ๐—–๐—ฎ๐—ฝ๐—ถ๐˜๐—ฎ๐—น ๐—ง๐—ต๐—ฒ ๐— ๐—ผ๐˜€๐˜ ๐—™๐—ฒ๐—ฎ๐—ฟ๐—ฒ๐—ฑ ๐—ฃ๐—ฟ๐—ถ๐—ป๐˜ ๐—ผ๐—ณ ๐˜๐—ต๐—ฒ ๐—ฌ๐—ฒ๐—ฎ๐—ฟ $NVDA Nvidia reports FQ3 FY27 after the close today, its first print without Jensen Huang. ๐—ช๐—ต๐—ฒ๐—ฟ๐—ฒ ๐—ฆ๐—ฒ๐—ป๐˜๐—ถ๐—บ๐—ฒ๐—ป๐˜ ๐—ฆ๐˜๐—ฎ๐—ป๐—ฑ๐˜€ โ€ข As of Monday's close, NVDA sits at $156.14, its lowest close since July 2025 and about 36% below the Oct 6 record of $243.37 โ€ข About 11x this year's consensus earnings, down from 17x last quarter โ€ข With sentiment this extreme and this low, any surprise will provide genuine relief to the AI trade โ€ข In my view this is a multiple reset, not a demand problem. Demand still outruns supply, and the real limits are memory, power and financing ๐—ง๐—ต๐—ฒ ๐—ฆ๐—ฒ๐˜๐˜‚๐—ฝ โ€ข Guide was $108.0B, plus or minus 2%, at a 74.0% gross margin. Consensus sits near $109B and $2.47 to $2.48 in EPS โ€ข The number that matters is the Q4 guide, with consensus near $125B โ€ข I believe NVIDIA may stick to guidance. The alternate is a beat and a guide down, which could cause severe shock to the trade ๐—Ÿ๐—ถ๐—ณ๐—ฒ ๐—”๐—ณ๐˜๐—ฒ๐—ฟ ๐—๐—ฒ๐—ป๐˜€๐—ฒ๐—ป โ€ข The team can run this company without him. Vera Rubin is ramping into full production and the roadmap ahead is set โ€ข What is gone is the lean structure. Jensen ran roughly 60 direct reports with few layers in between. Someone now has to orchestrate, and that likely means more hierarchy and slower decisions โ€ข Colette Kress, CFO since 2013 and now interim CEO, enabled the company to keep investing all these years. On the last call she was the one who laid out the neocloud revenue sharing model. Why would she hinder it now? โ€ข Frankly, adding some internal controls from a risk perspective will only legitimize their position ๐—ข๐˜‚๐—ฟ ๐—ฃ๐—ผ๐˜€๐—ถ๐˜๐—ถ๐—ผ๐—ป โ€ข We are holding off on a purchase until we have clarity on the CEO search and guidance "Buy when others are fearful" could not be more true with the way things stand right now.
1
1
1
3
2,031
๐—ก๐—ฒ๐—ฏ๐—ถ๐˜‚๐˜€ (๐—ก๐—•๐—œ๐—ฆ): ๐—–๐—น๐—ฒ๐—ฎ๐—ป ๐—•๐—ฒ๐—ฎ๐˜, ๐Ÿฎ๐Ÿฌ๐Ÿฎ๐Ÿณ ๐—ง๐—ฟ๐—ฎ๐—ท๐—ฒ๐—ฐ๐˜๐—ผ๐—ฟ๐˜† ๐—œ๐—ป๐˜๐—ฎ๐—ฐ๐˜. ๐—•๐˜‚๐˜†, ๐—ฃ๐—ง $๐Ÿฎ๐Ÿฑ๐Ÿฌ We reiterate Buy on $NBIS with a $250 PT. Q3 beat on the top line and on profitability, confirming demand for dedicated AI capacity remains strong. FY26 lands at the low end of the range on timing, not demand, and the new 2027 outlook keeps our thesis intact. ๐—ฅ๐—ฒ๐˜ƒ๐—ฒ๐—ป๐˜‚๐—ฒ โ€ข $980.8M vs consensus of about $907.5M, an 8% beat. โ€ข Up 571% YoY and 68% QoQ from $582.3M in Q2. โ€ข AI Cloud was $973.8M, up 643% YoY and about 99% of revenue. โ€ข Includes about $325M of non-recurring revenue from the first phase of the asset-light model. ๐—ฃ๐—ฟ๐—ผ๐—ณ๐—ถ๐˜๐—ฎ๐—ฏ๐—ถ๐—น๐—ถ๐˜๐˜† โ€ข Adjusted EBITDA of $465.8M, well ahead of expectations in the mid $350M range. โ€ข AI Cloud adjusted EBITDA margin of 53.2%. โ€ข Adjusted net income of $90.9M vs an expected loss. โ€ข Operating profit turned positive at $50.4M. The GAAP net loss of $37.5M, or $(0.13) per share, reflects interest and a one-time convertible note charge. ๐—™๐—ผ๐—ฟ๐˜„๐—ฎ๐—ฟ๐—ฑ ๐—ถ๐—ป๐—ฑ๐—ถ๐—ฐ๐—ฎ๐˜๐—ผ๐—ฟ๐˜€ โ€ข 2027 revenue guide of $12B to $13B, roughly 4x 2026, which management calls conservative. โ€ข ARR reached $6.4B at the end of September, up 113% from $3.0B in June. โ€ข $12B of TCV signed since the start of Q3. โ€ข ACV per MW is now above $25M, up from $20M to $25M. โ€ข Prepayments now cover more than 100% of related capex on relevant contracts, up from 50% to 60%. โ€ข Pre-tax breakeven targeted by the end of 2027. ๐—ก๐—ผ๐˜„, ๐—น๐—ฒ๐˜'๐˜€ ๐˜๐—ฎ๐—น๐—ธ ๐—ฎ๐—ฏ๐—ผ๐˜‚๐˜ ๐˜๐—ต๐—ฒ ๐—ฒ๐—น๐—ฒ๐—ฝ๐—ต๐—ฎ๐—ป๐˜ ๐—ถ๐—ป ๐˜๐—ต๐—ฒ ๐—ฟ๐—ผ๐—ผ๐—บ โ€ข FY26 revenue now tracks the low end of $3.0B to $3.4B, vs a Street midpoint near $3.35B. Some third-party capacity slipped into early 2027. โ€ข The $7B ARR milestone moves from December to February. โ€ข 2027 capex is guided at $40B to $45B, with 75% to 85% covered by customer prepayments and asset-backed financing. ๐—•๐—น๐—ผ๐—ผ๐—บ ๐—˜๐—ป๐—ฒ๐—ฟ๐—ด๐˜† ๐—ถ๐˜€ ๐—ฝ๐—ผ๐˜„๐—ฒ๐—ฟ๐—ถ๐—ป๐—ด ๐—ก๐—ฒ๐˜„ ๐—๐—ฒ๐—ฟ๐˜€๐—ฒ๐˜† โ€ข Nebius is moving its New Jersey site, which serves Microsoft, to Bloom Energy fuel cells as the main power source. โ€ข In the shareholder letter: "Bloom generation is already online and we are now scaling this capacity, expecting to deliver additional capacity over the next few weeks." โ€ข Nebius expects more than half of the Microsoft capacity in New Jersey delivered by year end, with the remaining tranches in 2Q27. โ€ข We read this as a validation of our Buy on $BE. Energy is the bottleneck, and on-site fuel cells are how neoclouds get capacity online. โ€ข The New Jersey slip came as temporary generation was retired ahead of the move to permanent Bloom power, resetting the schedule for the remaining tranches. ๐—ข๐˜‚๐—ฟ ๐˜ƒ๐—ถ๐—ฒ๐˜„ This is a timing slip, not a demand problem. Customers are signing longer contracts at higher prices with larger upfront payments. The $12B to $13B 2027 starting point, strong EBITDA conversion and prepayments that de-risk the capex ramp are the numbers that matter. We remain Buy with a $250 PT. INH Capital invests its own capital. We publish the research behind it. We are not an investment adviser and do not manage client funds. This is not investment advice.
$NBIS ๐—ก๐—•๐—œ๐—ฆ ๐—ค๐Ÿฏ ๐—ฃ๐—ฟ๐—ฒ๐˜ƒ๐—ถ๐—ฒ๐˜„: ๐—ฅ๐—ฎ๐—ถ๐˜€๐—ถ๐—ป๐—ด ๐—ฃ๐—ง ๐˜๐—ผ $๐Ÿฎ๐Ÿฑ๐Ÿฌ ๐—”๐—ต๐—ฒ๐—ฎ๐—ฑ ๐—ผ๐—ณ ๐—˜๐—ฎ๐—ฟ๐—ป๐—ถ๐—ป๐—ด๐˜€ Nebius reports Thursday, Nov 12th, before the open (call 8:00 AM ET). We rate Buy and raise our PT to $250 from $225. We expect revenue above consensus and the year-end ARR guide to hold or rise. We remain constructive on the market. We believe it prices in worst-case outcomes that we do not expect to materialize. CoreWeave's Q3 results, reported Nov 10, solidify the neocloud story and narrative. We believe Nebius has the better management and the better backlog. First, let's start with what the CoreWeave print told us โ€ข Revenue was $3.661B, up 168% YoY and up more than $1B sequentially. Adjusted EBITDA was $2.256B, a 62% margin. Adjusted operating income was $379M. โ€ข Revenue backlog was about $132.5B as of Sept 30, up from $104B at the end of June. โ€ข Full-year 2026 guidance was raised: revenue $13.0B to $13.3B, adjusted operating income $1.275B to $1.375B, and exit ARR $20B to $21B. On the backlog comparison โ€ข These figures are not like-for-like. Nebius has disclosed about $40B of customer commitments, as of its Aug 12 Q2 letter. That is not reported as remaining performance obligations, and it has not been updated for Q3. โ€ข CoreWeave's disclosed revenue backlog is larger in dollars and is a quarter more current. We are not claiming the Nebius number is bigger. Our view is on management and on the quality of the book. This quarter at Nebius โ€ข Consensus revenue is $907M, which would be up 56% from Q2 and about 521% from year-ago revenue of $146.1M. Consensus Q4 revenue is $1.46B, which would put the year near the top of the $3.0B to $3.4B guide. โ€ข Our estimates: Q3 revenue $940M, ARR exiting Q3 of $4.5B or more, and FY26 revenue at or above the $3.4B top of the guide. โ€ข Q2 revenue was $582.3M, up 454% YoY, with AI Cloud ARR of $3.0B. The 2026 ARR, revenue, margin and capex guide was reaffirmed. Contracted power target is 5 GW. Connected power guide is 800 MW to 1 GW. The setup โ€ข $NBIS closed at $134.00 on Nov 10, 55% below its $299.86 52-week high on June 22. The shares are still up about 60% year to date. โ€ข Our raised $250 PT implies 87% upside from the Nov 10 close, versus 68% at the prior $225 target. We believe the drawdown, the rebound that follows, and a firmer neocloud narrative will carry the shares to our new target. Now, let's talk about the elephant in the room โ€ข The ARR guide still requires a steep second-half ramp, from $3.0B at the end of June to $7B to $9B by year-end. โ€ข Capex of $20B to $25B is several times 2026 revenue, and funding still depends on prepayments, debt and equity, including the ATM. โ€ข Commitments are concentrated. Microsoft and Meta are the bulk of the $40B book, the $15B Meta option is contingent, and newly signed mid-term deals mostly contribute in late 2026 and 2027, not this quarter. โ€ข Connected power does not become revenue immediately. Management has said commissioning, cluster build and onboarding can take several months. The neocloud demand case just got a cleaner read. We prefer Nebius on management and on the quality of the backlog we can see. We stay constructive, keep the Buy, and raise the PT to $250 from $225 ahead of the print. INH Capital invests its own capital. We publish the research behind it. We are not an investment adviser and do not manage client funds. Not investment advice.
1
1
3
1,562
Buy-Side Note | $ONDS Prints a Clean Beat and Raises the Bar โ€” Trajectory Intact and Accelerating Ondas delivered the print the market needed this morning: a clean revenue beat, heavy bookings conversion, and a higher multi-year outlook that keeps the steep ramp on track. Revenue vs. Consensus - Actual: $167.5M - Consensus: ~$148M (range clustered around $147โ€“148M) - Beat: ~13% - Sequential: +100% from $83.8M in Q2 - Year-over-year: >16x from $10.1M - Pro forma organic: +64% YoY This lands above the high end of managementโ€™s prior $140โ€“155M guide and continues the companyโ€™s consistent pattern of converting backlog into reported revenue. EPS GAAP EPS came in at $(0.10), better than the more conservative end of Street estimates (some published figures sat at โ€“$0.13 to โ€“$0.14). Adjusted EBITDA loss widened sequentially to $59.7M as the company continues to invest ahead of the expected Q4 ramp; management still expects operating-platform Adjusted EBITDA profitability in Q4 2026 and has pulled company-wide profitability forward to Q3 2027. Bookings, Backlog & Guidance - Captured $214M in new orders during the quarter - Backlog increased 36% sequentially to $835M - Cash, cash equivalents, restricted cash and short-term investments: $820.3M - Full-year 2026 revenue target raised to at least $550M (from the prior $525โ€“550M range). - Q4 targeted at > $250M with at least 75% pro forma organic year-over-year growth. - 2027 revenue target initiated at at least $1.1B, excluding additional acquisitions. Ondas Sentinel โ€” $100M Q3 Bookings Sentinel generated $100 million in Q3 bookings, underscoring its emergence as a core defense growth engine: - ULTRA: $53.7M in follow-on orders - IonStrike: $21.7M - Continued strength in Counter-UAS demand These wins sit inside a broader $214M quarterly bookings figure and support the thesis that Sentinel (anchored by ULTRA long-endurance ISR, IonStrike kinetic CUAS, and expanding CUAS capabilities) is scaling from product validation into meaningful program revenue. Margins & Path to Profitability Adjusted gross margin was 34.8% and Adjusted EBITDA loss was $59.7M as the company absorbs mix, early-stage ramp inefficiencies, and the cost of scaling. These are the expected costs of the buildout. The combination of a rising top line, a large and growing backlog, and the explicit timeline to operating-platform profitability keeps the leverage story intact. Bottom Line This is a high-conviction growth print. Revenue conversion beat expectations, bookings remain strong and diversified, the backlog is expanding, and management raised the bar on both 2026 and the newly initiated 2027 outlook. Sentinel is emerging as a meaningful defense platform, but the broader OAS and autonomous systems portfolio is converting demand at scale. The cash position provides runway, and the path to Adjusted EBITDA profitability is now closer. We remain bullish on the trajectory. PT $15.
$ONDS ๐—ข๐—ป๐—ฑ๐—ฎ๐˜€ (๐—ข๐—ก๐——๐—ฆ): ๐—œ๐—ป๐—ถ๐˜๐—ถ๐—ฎ๐˜๐—ถ๐—ป๐—ด ๐—ฎ๐˜ ๐—•๐˜‚๐˜†, ๐—ฃ๐—ง $๐Ÿญ๐Ÿฑ We initiate $ONDS at Buy with a 12-month PT of $15, 162% upside from the $5.73 close today, Nov 11th. This is prior to Q3 earnings results tomororw Thursday, Nov 12th, before the open (call 8:30 AM ET). We believe the stock prices in worst-case outcomes that we do not expect to materialize. ๐—ช๐—ต๐˜† ๐˜„๐—ฒ ๐—น๐—ถ๐—ธ๐—ฒ ๐—ถ๐˜ โ€ข Defense is the theme. Over $270M of orders were booked from June 30 through Oct 1, across counter-drone, ISR and precision strike. โ€ข Backlog went from $68M at year-end 2025 to $757M pro forma at June 30. โ€ข Q2 revenue was a record $83.8M, up 13x YoY. The FY26 target was raised to $525M to $550M. โ€ข $1.4B of cash and investments at June 30. ๐—ง๐—ต๐—ฒ ๐—ฃ๐—ฎ๐—น๐—ฎ๐—ป๐˜๐—ถ๐—ฟ ๐—ฟ๐—ฒ๐—น๐—ฎ๐˜๐—ถ๐—ผ๐—ป๐˜€๐—ต๐—ถ๐—ฝ ๐—ธ๐—ฒ๐—ฒ๐—ฝ๐˜€ ๐—ฑ๐—ฒ๐—ฒ๐—ฝ๐—ฒ๐—ป๐—ถ๐—ป๐—ด โ€ข Mar 2025: Palantir Foundry to scale Optimus and Iron Drone Raider operations. โ€ข Mar 2026: SkyWeaver ISR partnership with Palantir and World View. โ€ข Oct 12, 2026: Regna AI, powered by Palantir, with Ondas as its first customer. ๐—”๐—œ ๐—ถ๐—ป๐˜€๐—ถ๐—ฑ๐—ฒ ๐˜๐—ต๐—ฒ ๐—ฝ๐—ฟ๐—ผ๐—ฑ๐˜‚๐—ฐ๐˜ โ€ข In Ondas's own words, Iron Drone locks on to targets "using proprietary computer vision and artificial intelligence (AI) capabilities." ๐—ก๐—ผ๐˜„, ๐—น๐—ฒ๐˜'๐˜€ ๐˜๐—ฎ๐—น๐—ธ ๐—ฎ๐—ฏ๐—ผ๐˜‚๐˜ ๐˜๐—ต๐—ฒ ๐—ฒ๐—น๐—ฒ๐—ฝ๐—ต๐—ฎ๐—ป๐˜ ๐—ถ๐—ป ๐˜๐—ต๐—ฒ ๐—ฟ๐—ผ๐—ผ๐—บ โ€ข Dilution: about 571M shares outstanding after stock-funded deals. โ€ข Profitability: Q2 adjusted EBITDA loss was $50.6M. โ€ข Execution: the FY26 guide implies a Q4 of roughly $244M to $269M, about 3x the record quarter. Our view: the order book, Palantir and AI built into the product line support the Buy and the $15 PT. ๐—ค๐Ÿฏ ๐—™๐—ฌ๐Ÿฎ๐Ÿฒ ๐—ฝ๐—ฟ๐—ฒ๐˜ƒ๐—ถ๐—ฒ๐˜„ โ€ข Consensus revenue is $148.04M, inside management's $140M to $155M guide. Consensus EPS is -$0.13. Consensus EBITDA is about -$40M. โ€ข What we are watching: revenue vs consensus and the guide, a backlog update, whether the FY26 target is reaffirmed or raised, the implied Q4 ramp, cash and dilution, and Palantir/Regna updates. INH Capital invests its own capital. We publish the research behind it. We are not an investment adviser and do not manage client funds. This is not investment advice.
1
2
627
$ONDS ๐—ข๐—ป๐—ฑ๐—ฎ๐˜€ (๐—ข๐—ก๐——๐—ฆ): ๐—œ๐—ป๐—ถ๐˜๐—ถ๐—ฎ๐˜๐—ถ๐—ป๐—ด ๐—ฎ๐˜ ๐—•๐˜‚๐˜†, ๐—ฃ๐—ง $๐Ÿญ๐Ÿฑ We initiate $ONDS at Buy with a 12-month PT of $15, 162% upside from the $5.73 close today, Nov 11th. This is prior to Q3 earnings results tomororw Thursday, Nov 12th, before the open (call 8:30 AM ET). We believe the stock prices in worst-case outcomes that we do not expect to materialize. ๐—ช๐—ต๐˜† ๐˜„๐—ฒ ๐—น๐—ถ๐—ธ๐—ฒ ๐—ถ๐˜ โ€ข Defense is the theme. Over $270M of orders were booked from June 30 through Oct 1, across counter-drone, ISR and precision strike. โ€ข Backlog went from $68M at year-end 2025 to $757M pro forma at June 30. โ€ข Q2 revenue was a record $83.8M, up 13x YoY. The FY26 target was raised to $525M to $550M. โ€ข $1.4B of cash and investments at June 30. ๐—ง๐—ต๐—ฒ ๐—ฃ๐—ฎ๐—น๐—ฎ๐—ป๐˜๐—ถ๐—ฟ ๐—ฟ๐—ฒ๐—น๐—ฎ๐˜๐—ถ๐—ผ๐—ป๐˜€๐—ต๐—ถ๐—ฝ ๐—ธ๐—ฒ๐—ฒ๐—ฝ๐˜€ ๐—ฑ๐—ฒ๐—ฒ๐—ฝ๐—ฒ๐—ป๐—ถ๐—ป๐—ด โ€ข Mar 2025: Palantir Foundry to scale Optimus and Iron Drone Raider operations. โ€ข Mar 2026: SkyWeaver ISR partnership with Palantir and World View. โ€ข Oct 12, 2026: Regna AI, powered by Palantir, with Ondas as its first customer. ๐—”๐—œ ๐—ถ๐—ป๐˜€๐—ถ๐—ฑ๐—ฒ ๐˜๐—ต๐—ฒ ๐—ฝ๐—ฟ๐—ผ๐—ฑ๐˜‚๐—ฐ๐˜ โ€ข In Ondas's own words, Iron Drone locks on to targets "using proprietary computer vision and artificial intelligence (AI) capabilities." ๐—ก๐—ผ๐˜„, ๐—น๐—ฒ๐˜'๐˜€ ๐˜๐—ฎ๐—น๐—ธ ๐—ฎ๐—ฏ๐—ผ๐˜‚๐˜ ๐˜๐—ต๐—ฒ ๐—ฒ๐—น๐—ฒ๐—ฝ๐—ต๐—ฎ๐—ป๐˜ ๐—ถ๐—ป ๐˜๐—ต๐—ฒ ๐—ฟ๐—ผ๐—ผ๐—บ โ€ข Dilution: about 571M shares outstanding after stock-funded deals. โ€ข Profitability: Q2 adjusted EBITDA loss was $50.6M. โ€ข Execution: the FY26 guide implies a Q4 of roughly $244M to $269M, about 3x the record quarter. Our view: the order book, Palantir and AI built into the product line support the Buy and the $15 PT. ๐—ค๐Ÿฏ ๐—™๐—ฌ๐Ÿฎ๐Ÿฒ ๐—ฝ๐—ฟ๐—ฒ๐˜ƒ๐—ถ๐—ฒ๐˜„ โ€ข Consensus revenue is $148.04M, inside management's $140M to $155M guide. Consensus EPS is -$0.13. Consensus EBITDA is about -$40M. โ€ข What we are watching: revenue vs consensus and the guide, a backlog update, whether the FY26 target is reaffirmed or raised, the implied Q4 ramp, cash and dilution, and Palantir/Regna updates. INH Capital invests its own capital. We publish the research behind it. We are not an investment adviser and do not manage client funds. This is not investment advice.
2
1
7
1,975
$NBIS ๐—ก๐—•๐—œ๐—ฆ ๐—ค๐Ÿฏ ๐—ฃ๐—ฟ๐—ฒ๐˜ƒ๐—ถ๐—ฒ๐˜„: ๐—ฅ๐—ฎ๐—ถ๐˜€๐—ถ๐—ป๐—ด ๐—ฃ๐—ง ๐˜๐—ผ $๐Ÿฎ๐Ÿฑ๐Ÿฌ ๐—”๐—ต๐—ฒ๐—ฎ๐—ฑ ๐—ผ๐—ณ ๐—˜๐—ฎ๐—ฟ๐—ป๐—ถ๐—ป๐—ด๐˜€ Nebius reports Thursday, Nov 12th, before the open (call 8:00 AM ET). We rate Buy and raise our PT to $250 from $225. We expect revenue above consensus and the year-end ARR guide to hold or rise. We remain constructive on the market. We believe it prices in worst-case outcomes that we do not expect to materialize. CoreWeave's Q3 results, reported Nov 10, solidify the neocloud story and narrative. We believe Nebius has the better management and the better backlog. First, let's start with what the CoreWeave print told us โ€ข Revenue was $3.661B, up 168% YoY and up more than $1B sequentially. Adjusted EBITDA was $2.256B, a 62% margin. Adjusted operating income was $379M. โ€ข Revenue backlog was about $132.5B as of Sept 30, up from $104B at the end of June. โ€ข Full-year 2026 guidance was raised: revenue $13.0B to $13.3B, adjusted operating income $1.275B to $1.375B, and exit ARR $20B to $21B. On the backlog comparison โ€ข These figures are not like-for-like. Nebius has disclosed about $40B of customer commitments, as of its Aug 12 Q2 letter. That is not reported as remaining performance obligations, and it has not been updated for Q3. โ€ข CoreWeave's disclosed revenue backlog is larger in dollars and is a quarter more current. We are not claiming the Nebius number is bigger. Our view is on management and on the quality of the book. This quarter at Nebius โ€ข Consensus revenue is $907M, which would be up 56% from Q2 and about 521% from year-ago revenue of $146.1M. Consensus Q4 revenue is $1.46B, which would put the year near the top of the $3.0B to $3.4B guide. โ€ข Our estimates: Q3 revenue $940M, ARR exiting Q3 of $4.5B or more, and FY26 revenue at or above the $3.4B top of the guide. โ€ข Q2 revenue was $582.3M, up 454% YoY, with AI Cloud ARR of $3.0B. The 2026 ARR, revenue, margin and capex guide was reaffirmed. Contracted power target is 5 GW. Connected power guide is 800 MW to 1 GW. The setup โ€ข $NBIS closed at $134.00 on Nov 10, 55% below its $299.86 52-week high on June 22. The shares are still up about 60% year to date. โ€ข Our raised $250 PT implies 87% upside from the Nov 10 close, versus 68% at the prior $225 target. We believe the drawdown, the rebound that follows, and a firmer neocloud narrative will carry the shares to our new target. Now, let's talk about the elephant in the room โ€ข The ARR guide still requires a steep second-half ramp, from $3.0B at the end of June to $7B to $9B by year-end. โ€ข Capex of $20B to $25B is several times 2026 revenue, and funding still depends on prepayments, debt and equity, including the ATM. โ€ข Commitments are concentrated. Microsoft and Meta are the bulk of the $40B book, the $15B Meta option is contingent, and newly signed mid-term deals mostly contribute in late 2026 and 2027, not this quarter. โ€ข Connected power does not become revenue immediately. Management has said commissioning, cluster build and onboarding can take several months. The neocloud demand case just got a cleaner read. We prefer Nebius on management and on the quality of the backlog we can see. We stay constructive, keep the Buy, and raise the PT to $250 from $225 ahead of the print. INH Capital invests its own capital. We publish the research behind it. We are not an investment adviser and do not manage client funds. Not investment advice.
1
2
4
1,789
๐— ๐—ฎ๐—ฟ๐—ธ๐—ฒ๐˜ ๐—ฉ๐—ถ๐—ฒ๐˜„ (Week Ending 11/13): ๐—ง๐—ต๐—ฒ ๐—™๐—จ๐—— ๐—›๐—ฎ๐˜€ ๐—•๐—ฒ๐—ฒ๐—ป ๐—˜๐˜…๐—ต๐—ฎ๐˜‚๐˜€๐˜๐—ฒ๐—ฑ. ๐—ง๐—ต๐—ฒ๐—ฟ๐—ฒ ๐—œ๐˜€ ๐—ก๐—ผ ๐—•๐˜‚๐—ฏ๐—ฏ๐—น๐—ฒ. We remain constructive on the market. We believe it prices in worst-case outcomes that we do not expect to materialize, and the setup is now moving in our favor. First, let's start with the rout โ€ข The Nasdaq 100 fell 14.7% from its Oct 6 high to its Nov 5 low. The Nov 5 session alone was -6.4%. โ€ข Semis took the brunt. The $SOXX fell 38.4% from its June 22 high to Nov 5, and Nov 5 was its 2nd-worst day since 1994, behind only March 16, 2020. โ€ข In the second half through Nov 5: $INTC -62.6%, $LRCX -57.2%, $AMAT -55.8%, $MU -43.1%. Positioning and sentiment are washed out โ€ข At the Nov 5 low, the 14-day RSI on both the Nasdaq 100 and the SOX was in the most oversold 1% of their histories, the lowest since April 2025. โ€ข The Cboe total put/call ratio hit 1.06 on Nov 5, the highest reading of October and November. โ€ข CNN Fear & Greed sits at 33, in Fear. AAII bears outnumber bulls by 14 points. There is no bubble โ€ข The S&P 500 tech sector trades at 17.3x forward earnings, the lowest since March 2020 and below its 25-year average of 20.0x. โ€ข The S&P 500 is at 17.7x, its lowest in more than three years. โ€ข Bubbles do not trade at below-average multiples with sentiment in Fear. This is a repricing, not a collapse in fundamentals. The macro is turning โ€ข Brent settled at $91.49 on Nov 10, its lowest since Aug 31 and down 22% month to date, as the Iran war appears to be winding down. โ€ข Energy has been the primary driver of inflation. The CPI energy index rose 5.2% in September and slowed to 0.7% in October. โ€ข A resolution means lower oil, lower inflation expectations and lower yields. The 10-year sits at 4.98%. โ€ข The data is lagging. The Fed meets Dec 8-9, and November CPI isn't released until Dec 10. In our view, that argues for the Fed to hold and wait for data that reflects the drop in energy. A catalyst for risk appetite โ€ข Anthropic's IPO, targeting a valuation of about $1.4T, is already drawing orders for multiple times the shares on offer, per Bloomberg. Demand of that size for AI exposure points to a short-term euphoric setup into year-end. Now, let's talk about the elephant in the room โ€ข October CPI was hot: headline +0.4% vs 0.3% expected, core +0.3% for a third straight month. Not all of this is energy. โ€ข Markets expect a December hike, and Iran says no formal negotiations are underway. โ€ข The SOX is still about 32% below its high and below its 200-day moving average. The damage takes time to repair. Our view: the FUD has been exhausted. Sentiment is washed out, valuations are compressed and the macro is turning. We think the market has priced the hawkish path and has not priced a resolution. We stay constructive and keep buying weakness in semis and AI infrastructure. INH Capital invests its own capital. We publish the research behind it. We are not an investment adviser and do not manage client funds. Not investment advice.
1
1
2
970
๐—ก๐—ฎ๐˜€๐—ฑ๐—ฎ๐—พ ๐Ÿญ๐Ÿฌ๐Ÿฌ: ๐—ข๐—ป๐—ฒ-๐——๐—ฎ๐˜† ๐—ฅ๐—ฒ๐—ฝ๐—ฟ๐—ถ๐—ฐ๐—ถ๐—ป๐—ด ๐—ถ๐—ป ๐—–๐—ผ๐—ป๐˜๐—ฒ๐˜…๐˜ We remain constructive on the market. We believe it now prices in worst-case outcomes that we do not expect to materialize. The selloff in context: โ€จโ€ข The Nasdaq 100 fell 6.36% on November 5, its fourth-worst session since the start of 2020.โ€จโ€ข The index finished 14.7% below its prior closing peak, but still roughly three times its year-end 2019 level.โ€จโ€ข The magnitude of the decline deserves attention. It does not, by itself, establish that the long-term earnings outlook has deteriorated by the same amount. What we believe the market is overlookingโ€จโ€ข Semiconductor demand, compute investment and power infrastructure operate on multi-year cycles. We do not expect those investment cases to reverse because of one election.โ€จโ€ข Our view is that greater policy certainty can eventually outweigh the initial fear, provided corporate earnings and investment plans remain intact.โ€จโ€ข We see an opportunity in businesses whose share prices have fallen faster than our assessment of their earnings power. Constructive, not complacent โ€จโ€ข Lower prices do not automatically mean attractive valuations. Earnings revisions, cash generation and balance-sheet strength still matter.โ€จโ€ข Power availability, permitting and the economics of new data centers remain real constraints. Clear cost-sharing rules and additional generation would address those problems more directly than blanket construction bans.โ€จโ€ข We would reassess if weaker demand, falling returns on investment or sustained estimate cuts undermine the underlying thesis. We do not need an immediate rebound to remain bullish. We need businesses to keep executing. Our focus remains on that distinction.
2
1
2
384
Source: Yahoo Finance Nasdaq 100 closing data; INH Capital calculations through November 5, 2026. Historical comparison begins January 2020. Price returns exclude dividends. INH Capital invests its own capital. We publish the research behind it. We are not an investment adviser and do not manage client funds.
48
Semiconductor stocks delivered some of the strongest returns in the market in 1H 2026 โ€” then reversed sharply. $MU Micron led with +304.6%, followed by $INTC Intel (+278.4%), $MRVL Marvell (+251.0%), and $ARM Arm (+224.4%). Equipment names (Applied Materials, Lam, KLA) also posted triple-digit gains. $NVDA Nvidia, by contrast, rose just +7.4% in the first half โ€” the weakest in the group. Since June 30th, the picture has flipped. Most names are down 30 - 60%+ from those levels (Intel โ€“62.6%, $LRCX Lam โ€“57.2%, KLA โ€“56.9%, Marvell โ€“55.1%). Texas Instruments has held up best (โ€“5.7%). The result is a classic first-half melt-up followed by a second-half reset, with memory, foundry, and equipment stocks absorbing the bulk of the reversal. Data as of Nov. 5th regular-session close (INH Capital). INH CAPITAL ยท RESEARCH
1
1
1
2
2,127
๐— ๐—ฎ๐—ฟ๐—ธ๐—ฒ๐˜ ๐—ฉ๐—ถ๐—ฒ๐˜„: ๐—ง๐—ต๐—ฒ ๐—ฆ๐˜„๐—ฒ๐—ฒ๐—ฝ ๐—ช๐—ฎ๐˜€ ๐—˜๐˜…๐—ฝ๐—ฒ๐—ฐ๐˜๐—ฒ๐—ฑ. ๐—ก๐—ผ๐˜„ ๐—œ๐˜ ๐—œ๐˜€ ๐—–๐—ฒ๐—ฟ๐˜๐—ฎ๐—ถ๐—ป. We remain constructive on the market. We believe it now prices in worst-case outcomes that we do not expect to materialize. We think tonight's reaction to a Democratic sweep is too strong. ๐—ช๐—ต๐—ฎ๐˜ ๐—ฐ๐—ต๐—ฎ๐—ป๐—ด๐—ฒ๐—ฑ โ€จโ€ข Very little. The administration's stubbornness on the war hurt its chances for months, and the outcome was widely expected. โ€จโ€ข What changed tonight is certainty. A known outcome replaces an open question, and markets price certainty better than fear. โ€จโ€ข We expect certainty to outweigh fear as investors move from guessing to pricing a defined policy path. ๐—ช๐—ต๐˜† ๐˜„๐—ฒ ๐—ฑ๐—ผ ๐—ป๐—ผ๐˜ ๐—ฒ๐˜…๐—ฝ๐—ฒ๐—ฐ๐˜ ๐—ด๐—ฟ๐—ผ๐˜„๐˜๐—ต ๐˜๐—ผ ๐—ฏ๐—ฒ ๐—ต๐—ถ๐—ป๐—ฑ๐—ฒ๐—ฟ๐—ฒ๐—ฑ โ€จโ€ข Campaign rhetoric rarely turns into broad anti-growth law. Legislating takes time, compromise and votes that hold together. โ€จโ€ข Support for domestic semiconductor manufacturing has been bipartisan since the CHIPS Act. We see no reason that changes. โ€จโ€ข Keeping the US ahead in AI is a national competitiveness and security priority on both sides of the aisle. ๐—ข๐—ป ๐—ฐ๐—ฎ๐—น๐—น๐˜€ ๐—ณ๐—ผ๐—ฟ ๐—ฎ ๐—ฑ๐—ฎ๐˜๐—ฎ ๐—ฐ๐—ฒ๐—ป๐˜๐—ฒ๐—ฟ ๐—บ๐—ผ๐—ฟ๐—ฎ๐˜๐—ผ๐—ฟ๐—ถ๐˜‚๐—บ โ€จโ€ข Some Democrats have floated pausing new data center construction. We view this as drastic and unrealistic as policy. โ€จโ€ข Halting buildout does not remove demand for compute. It pushes the investment, the jobs and the tax base to other states and other countries. โ€จโ€ข The realistic path is managing the strain on the grid: requiring new sites to bring their own power, sharing transmission costs, and speeding up approval for new generation. โ€จโ€ข That path supports on-site and behind-the-meter power, which helps $BE and the energy side of the AI trade. โ€จโ€ข We expect the debate to shape how data centers get built, not whether they get built. ๐—ช๐—ต๐—ฒ๐—ฟ๐—ฒ ๐˜„๐—ฒ ๐˜€๐—ฒ๐—ฒ ๐—ฟ๐—ฒ๐—ฎ๐—น ๐—ฟ๐—ถ๐˜€๐—ธ๐˜€ โ€จWe are constructive, not complacent. These are the areas we are watching: โ€จโ€ข Taxes. A push to raise corporate tax rates or tighten buyback treatment would weigh on earnings at the margin. We see it as a multiple headwind, not a thesis breaker. โ€จโ€ข Antitrust and AI oversight. Closer scrutiny of the largest platforms is likely. It may slow deals and add compliance costs, but we do not expect it to slow capex on compute. โ€จโ€ข Power costs and local permitting. This is the most legitimate concern. Residential electricity bills are a real political issue, and we expect tougher local fights over siting. Operators that bring their own power will be in a better position. โ€จโ€ข Export controls. We expect controls on advanced chips to stay tight or get tighter. That is not new, and we believe it is largely in current estimates. โ€จโ€ข Fiscal path. Higher spending without offsets would put pressure on long-term rates, which matters most for long-duration growth multiples. ๐—ฅ๐—ฒ๐—ด๐˜‚๐—น๐—ฎ๐˜๐—ถ๐—ผ๐—ป ๐—ถ๐˜€ ๐—น๐—ฎ๐—ฟ๐—ด๐—ฒ๐—น๐˜† ๐—ฝ๐—ฟ๐—ถ๐—ฐ๐—ฒ๐—ฑ ๐—ถ๐—ป โ€จโ€ข None of these risks are new, and we believe most of them are already reflected in multiples. โ€จโ€ข Clear rules are better for capital planning than open-ended uncertainty. Hyperscalers plan capex over multiple years, not one election cycle. ๐—ง๐—ต๐—ฒ ๐˜€๐—ฒ๐—บ๐—ถ๐—ฐ๐—ผ๐—ป๐—ฑ๐˜‚๐—ฐ๐˜๐—ผ๐—ฟ ๐—ฎ๐—ป๐—ฑ ๐—”๐—œ ๐˜๐—ฟ๐—ฎ๐—ฑ๐—ฒ โ€จโ€ข Demand for compute, memory and power comes from buildouts that run for years, not from who controls Congress. โ€จโ€ข We do not expect the semiconductor or AI trade to change much over the coming years. โ€จโ€ข We view election-driven weakness as an opportunity in names where the fundamentals have not changed. ๐—ข๐˜‚๐—ฟ ๐—ฐ๐—ผ๐˜ƒ๐—ฒ๐—ฟ๐—ฎ๐—ด๐—ฒ โ€จโ€ข Buy: $AMD (PT $600), $MU (PT $2,000), $NBIS (PT $225), $BE (PT $250), $QS (PT $7)โ€จโ€ข Hold: $INTC (PT $80) Fear is loud tonight. We expect certainty to win out over the coming weeks. INH Capital invests its own capital. We publish the research behind it. We are not an investment adviser and do not manage client funds.
3
1
5
1,971
INH Capital: Intel ($INTC) Downgrade to Hold, PT $80 - Floundering Foundry We are downgrading Intel from 'Buy' to 'Hold' and cutting our price target from $175 to $80. Intel reported Q3 FY26 on Thursday, Oct 29th, after the close. Why $175 no longer holds โ€ข $175 was an execution call: continued growth, raised guidance, internal products on time, and substrate limits that were manageable with limited downside โ€ข Each assumption was tested on this print, and most did not hold โ€ข Q4 guidance of $15.0B to $16.0B revenue and $0.25 EPS vs consensus of $17.1B and $0.43. That was a shock โ€ข Q4 gross margin guided to 38.0% vs 44.1% in Q3 โ€ข Substrates moved from a manageable constraint to a limiter on server output, with relief weighted to the back half of 2027 โ€ข Free cash flow is now likely not positive in 2027, which management calls an investment year โ€ข Without execution, the multiple is not justified. $80 reflects the business as guided today Management communication โ€ข Communication was central to our thesis. It is now central to our downgrade โ€ข Recent progress language set the tone: 18A yields ahead of internal projections, "multiple potential customers" on 14A, and repeated "stay tuned" โ€ข The guide did not match that tone, and investors were not prepared for it โ€ข A near-term story is now a "multiyear journey." We feel that shift should have been communicated before the print โ€ข 14A still has no named customer and production is framed for 2028. Progress language without commitments makes the yield narrative hard to trust โ€ข Classiq was announced days before the print with no explanation of why it ranks ahead of substrate supply โ€ข The CFO's answers on free cash flow and capital did not address what investors needed to hear โ€ข Asked directly about 2027 free cash flow, the answer was that the number is not locked and is probably not positive. We heard no bridge back to positive cash generation and no funding priorities โ€ข On 14A and 18A-P customers the answer was again "stay tuned." โ€ข Supply was framed as improving, yet the first half of 2027 is still expected to be tight on wafers and substrates. After a $23B raise, we feel investors deserved a clearer answer on why supply was not secured earlier โ€ข In our view, the call explained the quarter but not the shortcomings. Management must address them plainly The Israel incident is forgiven โ€ข The Fab 28 accident takes roughly $300M of revenue and 200bp of gross margin out of Q4, mostly 10nm server parts, with a similar early revenue estimate for Q1 โ€ข We do not hold that against management. Our downgrade rests on pain points unique to Intel relative to its competitors Substrates and capital allocation โ€ข Management cited an industry-wide shortage of advanced organic substrates for EMIB packaging that is limiting server output โ€ข The first half of 2027 is expected to stay tight on wafers and substrates โ€ข 18A wafers in Oregon are being taken offline to add Intel 3 data center output, a reallocation we feel should have come sooner โ€ข Intel raised $23B in equity in August, then agreed to buy Classiq for a reported $2.3B in cash and stock, a company with revenue in the tens of millions โ€ข In our view, that capital could have secured substrate supply through prepayments Free cash flow โ€ข Q3 adjusted free cash flow was $1.6B, but 2027 is likely not positive โ€ข Gross capex is $20B in 2026 and guided up strongly in 2027 โ€ข Our view is that a near-term story has become a multi-year effort, with the company moving into cash burn 14A โ€ข The 14A 0.9 PDK shipped in October. There is still no named customer, and production is framed for 2028 โ€ข We do not yet trust the 14A yield narrative. We feel management has underplayed its execution issues What is working โ€ข DCAI revenue of $7.3B, +78% YoY, with server CPU demand at its strongest in 15 years โ€ข DCAI guided up mid-teens sequentially in Q4, with purpose-built silicon growing faster โ€ข Industry server CPU units expected to grow more than 30% in both 2026 and 2027, with server CPU TAM seen rising from $40B to more than $200B over five years โ€ข Clearwater Forest on 18A shipping for revenue, Diamond Rapids on 18A-P ramping in 2027, and Granite Rapids among the fastest ramping server products โ€ข DCAI operating profit of $3.0B, a 42% operating margin, up about $600M sequentially โ€ข Purpose-built silicon up more than 50% sequentially and more than double YoY, with management guiding it to grow faster than DCAI in Q4 โ€ข A couple of advanced packaging customers expected to be announced in the next two quarters, and the 14A 1.0 PDK targeted by Q1 โ€ข Multiple customers seeking long-term CPU supply agreements, alongside a new Oracle Cloud deployment โ€ข 18A costs down about 40% since Q1, with 18A now nearly a third of client CPU revenue โ€ข Q3 revenue of $16.5B (+21% YoY), 44.1% gross margin, and EPS of $0.52 vs $0.40 consensus Our view โ€ข The market reaction reads to us as a split verdict: a strong server story against a weak guide. We agree with both halves โ€ข On a 10-year horizon, we remain bullish on Intel โ€ข Today the thesis carries large 'ifs' on execution and on management's communication. A better-run Intel riding server demand is not yet a fixed Intel โ€ข What would change our view: a named external customer on 18A-P or 14A, Foundry losses narrowing meaningfully from $1.9B a quarter, and a credible path back to positive free cash flow โ€ข We feel the continued progress language has misled investors, and in our view there has been significant mismanagement. The answers on the call were insufficient โ€ข The drawdown, about 30% today and the worst day since 1974, is in our view too severe relative to the setback and may mark a capitulation point. We would rather wait for proof than chase it Hold. PT $80. INH Capital invests its own capital and publishes the research behind it. Opinions of INH Capital. Not investment advice.
New research from INH Capital Intel ($INTC): Finding Foundry Q3 FY26 earnings preview. Intel reports Thursday, Oct 29, after the close. Our estimates vs consensus โ€ข Q3 revenue: $17.6B vs $16.49B (+6.7%) โ€ข Q3 gross margin: 44.9% vs a 42.0% guide (+290bp) โ€ข Q3 EPS: $0.55 vs $0.40 (+38%) โ€ข Q4 EPS: $0.59 vs $0.44 โ€ข FY26 EPS: $1.85 vs $1.54 โ€ข FY27 EPS: $2.70 vs $2.18 What drives it โ€ข CPU price increases in July, with server demand still ahead of supply โ€ข Data center (DCAI) revenue of $7.6B, +85% YoY โ€ข Foundry funded by customers: prepays, advanced packaging and the 14A PDK 0.9 What matters Thursday Guidance and outlook often dominate. Stronger reactions have frequently hinged more on forward guidance, capital spending plans, foundry and yield progress, and AI and data center commentary than on the quarter's beat itself. Options imply a ~7.7% move as of Oct 28th. PT $175. The full note, with 11 sourced exhibits, is free: inhcapital.com/insights/reseโ€ฆ INH Capital invests its own capital and publishes the research behind it. Opinions of INH Capital. Not investment advice.
5
1
23
15,280
New research from INH Capital Intel ($INTC): Finding Foundry Q3 FY26 earnings preview. Intel reports Thursday, Oct 29, after the close. Our estimates vs consensus โ€ข Q3 revenue: $17.6B vs $16.49B (+6.7%) โ€ข Q3 gross margin: 44.9% vs a 42.0% guide (+290bp) โ€ข Q3 EPS: $0.55 vs $0.40 (+38%) โ€ข Q4 EPS: $0.59 vs $0.44 โ€ข FY26 EPS: $1.85 vs $1.54 โ€ข FY27 EPS: $2.70 vs $2.18 What drives it โ€ข CPU price increases in July, with server demand still ahead of supply โ€ข Data center (DCAI) revenue of $7.6B, +85% YoY โ€ข Foundry funded by customers: prepays, advanced packaging and the 14A PDK 0.9 What matters Thursday Guidance and outlook often dominate. Stronger reactions have frequently hinged more on forward guidance, capital spending plans, foundry and yield progress, and AI and data center commentary than on the quarter's beat itself. Options imply a ~7.7% move as of Oct 28th. PT $175. The full note, with 11 sourced exhibits, is free: inhcapital.com/insights/reseโ€ฆ INH Capital invests its own capital and publishes the research behind it. Opinions of INH Capital. Not investment advice.
8
1
48
6,597
The Dip with the Chips - INH Capital Weekly Friday Flex - Oct 23rd, 2026 The $SOXXโ€™s recent selloff is unusually sharp by streak length and dollar impact, but core earnings, backlog, and spend data remain intact for key names, consistent with a positioning / rates-driven washout rather than a demand break. The Philadelphia Semiconductor Index (SOX) logged eight consecutive down sessions through the Oct. 16th, 2026 close, its longest skid since 2008, for a -15.7% drop in that window and roughly -24% from the mid-June peak. Broader AI / semiconductor stocks shed an estimated $1-2.7 trillion in market value over the period. Selling continued into Oct. 22nd (SOX closed near 10,732, โ€“6.7% that day), with the group testing supports and the 200-day moving average. YTD the index remains solidly higher (47โ€“51%). Drivers include rates sensitivity, higher oil (WTI near or above $100, Brent ~$115 amid Middle East tensions), financing anxiety around leveraged AI infrastructure (e.g., CoreWeave CDS and debt headlines), ROI skepticism on hyperscaler spend, power-grid constraints, and a positioning unwind. Hyperscaler 2026 capex guidance remains in the ~$700-800 billion range (with some estimates higher), and many companies continue to describe being more supply-constrained than demand-constrained. Goldman Sachs prime-brokerage data confirms a rapid de-risking of the crowded AI / tech trade. Hedge funds cut technology holdings at the fastest percentage pace in records dating to at least 2016 over the five sessions through Oct. 15, with semiconductors leading dollar losses; this largely reversed post-2025 net AI buying via long liquidation rather than new shorts. Fundamental long-short funds lost ~2.9% that week (still positive YTD). On Oct. 22nd, all three major equity strategies tracked by Goldman (fundamental L/S, systematic, multi-strategy) fell more than 1% in a single day, the first such breadth since the COVID period, with tech-focused funds down roughly 10% month-to-date through that date and Asia-focused managers also hard hit. Net leverage dropped to a one-year low, though tech still represented a high historical share of net exposure. This aligns with a sentiment / positioning washout more than a fundamental earnings collapse. On the specific names and valuations: $MU Micronโ€™s fiscal Q3 2026 results (reported late September) were exceptional - revenue $54.2 billion (+379% YoY), non-GAAP EPS $33.42, with a backlog cited near $100 billion and a large majority of 2027 output (including essentially all planned HBM) already contracted at higher prices. Forward multiples have compressed sharply (some quotes around ~5x), well below historical averages even as margins and visibility remain strong. $INTC Intel has shown tangible foundry progress (18A output exceeding internal targets, High-NA EUV in high-volume manufacturing, tool installation at Fab 62 targeting late-2028 operations, and named customer engagement including Tesla on 14A), yet the stock has drawn down substantially (recent closes near $85 from levels above $120 earlier in October), improving the asymmetric setup on continued execution. $NBIS Nebius continues to target strong revenue/ARR growth ($3-3.4 billion 2026 revenue guidance and higher year-end run-rate) amid AI compute shortages, but it carries higher leverage and CapEx intensity ($20-25 billion guided), so financing noise hits it harder; it has pulled back substantially from June highs. Broader AI-infrastructure valuations have de-rated (memory multiples in particular embed significant pessimism relative to current margins and contracted demand), while hyperscaler spend guidance has not been walked back in a way that breaks the near-term demand signal. Risks (rates path into midterms, oil-driven inflation, financing for highly leveraged players, ROI on the buildout, and competitive noise) are real and have not fully cleared, but the market is pricing them as more advanced than the current order books and earnings trajectories indicate. A sustained easing in oil / commodity strength or yields would likely support a sharp snap-back across the complex given the crowded unwind already completed. This is not investment advice. The sector remains highly volatile and sensitive to macro and financing headlines.
3
1
1
5
1,551
INH Capital Update โ€” We Are Buying the Dip The recent multi-session selloff across semis and AI infrastructure - 8 consecutive down days in the SOX, the longest streak since 2008, including Fridayโ€™s action - has pushed several of our core holdings into levels we view as overly discounted relative to unchanged fundamental trajectories. Sentiment, rates sensitivity, and positioning have driven the move. Business outcomes have not. Much of the incremental noise (debt headlines, competitive claims, and financing fears) has been manufactured or amplified well beyond the underlying facts. We are ignoring it where the data do not support a thesis change. Rates remain the primary external variable. Higher oil and sticky inflation prints have reopened questions around the path of policy into the midterms, with markets pricing a more cautious or delayed easing cycle. That backdrop raises the cost of capital for leveraged AI infrastructure plays and has amplified concerns around financing the buildout. We do not dismiss the rate risk, but we also do not see it as sufficient to break the demand signal. Hyperscaler capital expenditure remains on track for the $700 billion range in 2026, companies continue to report being supply-constrained rather than demand-constrained, and multi-year contracted backlogs at neoclouds more than cover near-term obligations. Nvidia residual capacity support further limits the downside case. The market has been trying to get ahead of a financing or demand break that has not materialized and, in our view, is unlikely to materialize in the near term given the scale of committed inflows. The sharp widening in credit-default-swap spreads on leveraged AI infrastructure names ($CRWV CoreWeaveโ€™s five-year CDS briefly above 1,100 bps) is, in our view, an overblown reflection of near-term rate and financing anxiety rather than a fundamental break. While current borrowing costs are elevated and the near-term rate path remains biased higher into the midterms, we continue to see the demand and backlog picture as intact. We are disregarding the extreme CDS readings as a reason to abandon the thesis. Competitive noise from CXMTโ€™s 4Fยฒ server-DRAM claims has added to the sentiment washout. Those advances are relevant for Chinaโ€™s domestic market, but US companies cannot purchase advanced product from CXMT under the existing export-control and 1260H framework, and the risk of further restrictions remains elevated. This does not reopen the Western supply valve. Memory remains structurally constrained rather than cyclical. We reiterate our top buys. $MU โ€” Maintain Buy. Micron continues to benefit from the HBM and high-end DRAM supercycle. The large majority of 2027 output (including essentially all planned HBM capacity) is already contracted at higher prices, with industry tightness expected to intensify into 2027โ€“2028 as manufacturers prioritize AI-grade product. Long-term customer agreements provide visibility that was absent in prior cycles. Nothing in the fundamental setup has deteriorated; the stock has simply been caught in the broader semiconductor de-risking. Risk/reward is more attractive at current levels. $NBIS โ€” Maintain Buy. Nebius has been hit alongside other AI infrastructure and neocloud names on debt and capex concerns. The contracted power and capacity story is unchanged, and dedicated AI compute demand remains intact. The sharp pullback improves the entry point for a name that can still deliver significant ARR growth as capacity comes online. The financing noise is real in a higher-rate environment, but it does not alter the backlog or the structural shortage of AI compute. We remain constructive and are adding on weakness. Anyone mapping the AI supply chain understands that power, advanced packaging, HBM, and leading-edge logic continue to bottleneck frontier model training and the path toward AGI/ASI-scale systems. Policy uncertainty into the midterms is real. Potential shifts in fiscal stance, trade, and tech regulation can affect multiples. However, it does not change the underlying chip, buildout, or demand story. The trade was overly sold on positioning and rates fears. Across the group we see no material deterioration in business trajectories. We continue to view $INTC Intel as the highest-conviction asymmetric setup. Foundry progress on 18A (wafer output, 18A-P in risk production) and 14A customer engagement remains intact, while the stock has absorbed significant size. The risk/reward has improved materially from recent highs, provided the macro rates path stays within a range that does not force a broad de-risking of the AI complex. We are using the window and buying or adding where the thesis is intact and technical exhaustion is extreme. Not financial advice.
3
1
9
1,856
INH Capital retweeted
INH Capital Update โ€” Using the Drawdown as an Opportunity The recent multi-session selloff across semis and AI infrastructure has pushed several of our core holdings into levels we view as overly discounted relative to unchanged fundamental trajectories. Sentiment, rates sensitivity, and positioning have driven the move. Business outcomes have not. We are using the window. $INTC โ€” Maintain Buy. Price target raised to $170 from $150. Intel remains one of the cleaner asymmetric setups in the group. Data Center & AI growth, foundry progress (18A ramp with improving yields, 18A-P in risk production, 14A on track for H2 2027 risk production), and advanced packaging optionality are intact. The stock has absorbed size on the way down and is now trading with daily oscillators in extreme oversold territory and a weekly Ulcer Index near 25. We continue to add on weakness. $170 reflects longer-term EPS power and foundry contribution that the market is still underwriting conservatively. Risk/reward has improved materially from the recent highs. $AMD โ€” Maintain Hold. Price target raised to $600 from $500. AMDโ€™s growth profile in data center GPUs and EPYC remains strong, and the multiple expansion that accompanied it has now partially reversed. We are not chasing the name aggressively here, but the drawdown improves the setup enough that we are lifting the target. The business is executing; the stock simply got ahead of near-term digestion. We hold the position and will reassess on further weakness or clear evidence of sustained share gains. $NBIS โ€” Upgrade to Buy. Price target raised to $300 from $250. Nebius has been hit alongside other AI infrastructure and neocloud names on debt and capex concerns. The contracted power and capacity story is unchanged, and dedicated AI compute demand remains intact. The stockโ€™s sharp pullback from the mid-to-high 200s improves the entry point for a name that can still deliver significant ARR growth as capacity comes online. We are upgrading and adding. $300 reflects the longer-term economics of owned power and scaled GPU clusters once the current sentiment washout fades. $MU โ€” Maintain Buy. Price target raised to $1750 from $1000. Micron continues to benefit from the HBM and high-end DRAM supercycle. Supply remains constrained, pricing power is elevated, and long-term customer agreements provide visibility that was not present in prior cycles. Nothing in the fundamental setup has deteriorated; the stock has simply been caught in the broader semiconductor de-risking. We maintain the Buy and are raising the target to reflect both the improved entry and continued tightness into 2027โ€“2028. Risk/reward is more attractive at current levels than it was two weeks ago. Across the group, we see no material change in the underlying businesses. The drawdown is primarily sentiment- and positioning-driven. That improves the risk/reward on names where we already had high conviction. We are buying or adding where the thesis is intact and the technical exhaustion is extreme. Not financial advice.
3
3
423
INH Capital Update โ€” Using the Drawdown as an Opportunity The recent multi-session selloff across semis and AI infrastructure has pushed several of our core holdings into levels we view as overly discounted relative to unchanged fundamental trajectories. Sentiment, rates sensitivity, and positioning have driven the move. Business outcomes have not. We are using the window. $INTC โ€” Maintain Buy. Price target raised to $170 from $150. Intel remains one of the cleaner asymmetric setups in the group. Data Center & AI growth, foundry progress (18A ramp with improving yields, 18A-P in risk production, 14A on track for H2 2027 risk production), and advanced packaging optionality are intact. The stock has absorbed size on the way down and is now trading with daily oscillators in extreme oversold territory and a weekly Ulcer Index near 25. We continue to add on weakness. $170 reflects longer-term EPS power and foundry contribution that the market is still underwriting conservatively. Risk/reward has improved materially from the recent highs. $AMD โ€” Maintain Hold. Price target raised to $600 from $500. AMDโ€™s growth profile in data center GPUs and EPYC remains strong, and the multiple expansion that accompanied it has now partially reversed. We are not chasing the name aggressively here, but the drawdown improves the setup enough that we are lifting the target. The business is executing; the stock simply got ahead of near-term digestion. We hold the position and will reassess on further weakness or clear evidence of sustained share gains. $NBIS โ€” Upgrade to Buy. Price target raised to $300 from $250. Nebius has been hit alongside other AI infrastructure and neocloud names on debt and capex concerns. The contracted power and capacity story is unchanged, and dedicated AI compute demand remains intact. The stockโ€™s sharp pullback from the mid-to-high 200s improves the entry point for a name that can still deliver significant ARR growth as capacity comes online. We are upgrading and adding. $300 reflects the longer-term economics of owned power and scaled GPU clusters once the current sentiment washout fades. $MU โ€” Maintain Buy. Price target raised to $1750 from $1000. Micron continues to benefit from the HBM and high-end DRAM supercycle. Supply remains constrained, pricing power is elevated, and long-term customer agreements provide visibility that was not present in prior cycles. Nothing in the fundamental setup has deteriorated; the stock has simply been caught in the broader semiconductor de-risking. We maintain the Buy and are raising the target to reflect both the improved entry and continued tightness into 2027โ€“2028. Risk/reward is more attractive at current levels than it was two weeks ago. Across the group, we see no material change in the underlying businesses. The drawdown is primarily sentiment- and positioning-driven. That improves the risk/reward on names where we already had high conviction. We are buying or adding where the thesis is intact and the technical exhaustion is extreme. Not financial advice.
3
3
423
INH Capital retweeted
๐—•๐˜‚๐˜†-๐—ฆ๐—ถ๐—ฑ๐—ฒ ๐—ก๐—ผ๐˜๐—ฒ | ๐—–๐—ผ๐—ป๐˜€๐˜๐—ฟ๐˜‚๐—ฐ๐˜๐—ถ๐˜ƒ๐—ฒ ๐—ง๐—ต๐—ฟ๐—ผ๐˜‚๐—ด๐—ต ๐˜๐—ต๐—ฒ ๐—”๐—œ ๐—–๐—ผ๐—บ๐—ฝ๐—น๐—ฒ๐˜… ๐——๐—ฟ๐—ฎ๐˜„๐—ฑ๐—ผ๐˜„๐—ป - ๐—œ๐—ก๐—› ๐—–๐—ฎ๐—ฝ๐—ถ๐˜๐—ฎ๐—น We remain bullish on the AI infrastructure and semiconductor complex through the current drawdown. The move has been driven primarily by elevated risk aversion - hot inflation prints raising the probability of further Fed tightening, concerns around the financing and capital intensity of the neocloud buildout, and broader โ€œAI tradeโ€ positioning unwind - rather than a deterioration in underlying demand for compute, memory, or advanced packaging. Macro setup: We continue to expect the broader environment to improve. While recent CPI data has shifted near-term rate pricing, our base case is that the Fed holds the policy rate and does not embark on a sustained hiking cycle. Upcoming PPI will be informative, but the subsequent PCE print is the more important confirmation point for the path of policy and financial conditions. We are using the current window of higher uncertainty and weaker sentiment to add to high-conviction positions. $NBIS (Nebius) โ€จHigh-growth AI-native cloud / neocloud exposure with substantial contracted power and capacity. The stock has been hit hard alongside peers on debt and capex concerns, but we view the demand backdrop for dedicated AI compute as intact and the long-term economics of owned power and scaled GPU clusters as attractive. Recent weakness improves the entry point for a name that can still deliver significant revenue and ARR growth as capacity comes online. $INTC (Intel) โ€จStrong Data Center & AI growth, improving gross margins, and progress across both product and manufacturing. Foundry note: Based on sources we track, we identify the foundry roadmap as strong going into H1 2027. 18A is ramping with improving yields; 18A-P is in risk production with meaningful performance/power gains; and 14A remains on track for risk production in H2 2027 (with volume in 2028), supported by defect-density progress, updated PDKs, and rising customer evaluation activity. Advanced packaging (EMIB/Foveros) is also scaling to support larger AI systems, with key elements targeting 2027 production readiness. Narrowing foundry losses and a clearer path toward breakeven around 2027 reinforce the multi-year recovery case. We continue to see Intel as a recovery story that benefits from both CPU demand in AI servers and the shift toward more diverse leading-edge and packaging capacity. The risk/reward remains favorable on further pullbacks. $MU (Micron) โ€จDirect beneficiary of the HBM and high-end DRAM supercycle. Supply remains tight (management has signaled conditions could stay constrained through 2027โ€“2028), pricing power is elevated, and long-term customer agreements provide visibility. Memory is the critical bottleneck in AI systems; we expect this to remain a multi-year structural tailwind rather than a one-cycle event. $ARM (Arm Holdings) โ€” initiated โ€จWe initiated a position in Arm alongside our Intel exposure. Arm is capturing rising royalty and licensing opportunity from AI data-center CPUs (efficiency, agentic workloads, custom silicon). Pairing it with Intel gives complementary exposure across x86/foundry and Arm architecture royalties. At current levels after the broader drawdown, the risk/reward is attractive relative to the multi-year AI infrastructure buildout. Positioning & risks: We are adding selectively rather than aggressively. Key risks we are monitoring: a sustained re-acceleration in inflation that forces additional Fed hikes, further stress in highly leveraged AI infrastructure financing, and any signs of slowing hyperscaler capex or memory demand. None of these currently alter our constructive medium-term view. We will continue to buy through the noise while the fundamental AI infrastructure and memory theses remain intact.
4
2
5
4,148
๐—ฃ๐—ฃ๐—œ ๐—˜๐—ป๐—ฒ๐—ฟ๐—ด๐˜† ๐—ฆ๐—ต๐—ผ๐—ฐ๐—ธ & ๐—œ๐—ป๐˜๐—ฟ๐—ฎ๐—ฑ๐—ฎ๐˜† ๐—ฆ๐˜๐—ฎ๐—ฏ๐—ถ๐—น๐—ถ๐˜‡๐—ฎ๐˜๐—ถ๐—ผ๐—ป - ๐—–๐—ผ๐—ป๐˜€๐˜๐—ฟ๐˜‚๐—ฐ๐˜๐—ถ๐˜ƒ๐—ฒ ๐—ผ๐—ป ๐˜๐—ต๐—ฒ ๐—–๐—ผ๐—บ๐—ฝ๐—น๐—ฒ๐˜… โ€” ๐—œ๐—ก๐—› ๐—–๐—ฎ๐—ฝ๐—ถ๐˜๐—ฎ๐—น The September PPI printed far hotter than expected (+1.3% m/m vs. ~0.5โ€“0.6% consensus, +6.3% y/y), the largest monthly gain since March 2022 and the sharpest annual rise since late 2022. Final-demand goods jumped 2.6% (the largest increase since the series began in 2009), with energy +10.8% accounting for more than 80% of that rise and gasoline alone (+20.3%) driving roughly half of the goods advance. Services rose a more moderate 0.8%. The core measure excluding food, energy, and trade services advanced 0.5% m/m (+5.0% y/y), confirming some broadening but still consistent with the energy impulse as the dominant near-term driver rather than a broad-based reacceleration in underlying pricing power. This follows yesterdayโ€™s CPI, which was also energy-led. The temporary nature of the supply shock (tied to Middle East hostilities and the associated crude/gasoline spike) remains the base case for the path of PCE - the Fedโ€™s preferred gauge. Estimates for September core PCE have moved higher but continue to point to a more contained reading once the energy pass-through is isolated; the subsequent PCE release will be the cleaner confirmation for policy and financial conditions. While the hot PPI raises the probability that the Fed feels pressure to deliver another hike (October odds have fluctuated and remain data-dependent), a sustained multi-hike cycle is still not our base case if energy effects fade and the labor market stays balanced. In this environment, the investable set remains constrained. Real rates and policy uncertainty favor assets with durable cash-flow growth over pure duration or speculative beta. High-conviction positions in the AI infrastructure and semiconductor complex continue to screen favorably on that basis: earnings power is still expanding with contracted demand, tight memory supply, and improving product/foundry execution, even as positioning and financing concerns have driven the recent drawdown. Price action today reinforced the constructive setup. Multiple semiconductor names sold off hard from the open on the PPI print and residual AI-complex risk aversion, then stabilized and closed well off their intraday lows, with several printing hammer-style candles (strong lower wicks and recovery into the close). Micron finished green on the day, Arm closed higher, and names including Intel and others recovered meaningfully from session lows even as the SOX finished modestly lower. This pattern of dip absorption is consistent with underlying demand remaining intact rather than a clean break in the fundamental thesis. We view the current window as one in which selective additions remain appropriate rather than a reason to reduce exposure. The medium-term theses on compute, HBM/DRAM, advanced packaging, and related CPU/foundry recovery are intact; near-term inflation noise and rate path uncertainty do not alter that assessment so long as hyperscaler and enterprise demand hold. Key risks we continue to monitor include a more persistent inflation impulse that forces a clearer hiking path, further stress in leveraged AI financing, or any evidence of slowing capex. None of these currently override the constructive view through the noise.
๐—•๐˜‚๐˜†-๐—ฆ๐—ถ๐—ฑ๐—ฒ ๐—ก๐—ผ๐˜๐—ฒ | ๐—–๐—ผ๐—ป๐˜€๐˜๐—ฟ๐˜‚๐—ฐ๐˜๐—ถ๐˜ƒ๐—ฒ ๐—ง๐—ต๐—ฟ๐—ผ๐˜‚๐—ด๐—ต ๐˜๐—ต๐—ฒ ๐—”๐—œ ๐—–๐—ผ๐—บ๐—ฝ๐—น๐—ฒ๐˜… ๐——๐—ฟ๐—ฎ๐˜„๐—ฑ๐—ผ๐˜„๐—ป - ๐—œ๐—ก๐—› ๐—–๐—ฎ๐—ฝ๐—ถ๐˜๐—ฎ๐—น We remain bullish on the AI infrastructure and semiconductor complex through the current drawdown. The move has been driven primarily by elevated risk aversion - hot inflation prints raising the probability of further Fed tightening, concerns around the financing and capital intensity of the neocloud buildout, and broader โ€œAI tradeโ€ positioning unwind - rather than a deterioration in underlying demand for compute, memory, or advanced packaging. Macro setup: We continue to expect the broader environment to improve. While recent CPI data has shifted near-term rate pricing, our base case is that the Fed holds the policy rate and does not embark on a sustained hiking cycle. Upcoming PPI will be informative, but the subsequent PCE print is the more important confirmation point for the path of policy and financial conditions. We are using the current window of higher uncertainty and weaker sentiment to add to high-conviction positions. $NBIS (Nebius) โ€จHigh-growth AI-native cloud / neocloud exposure with substantial contracted power and capacity. The stock has been hit hard alongside peers on debt and capex concerns, but we view the demand backdrop for dedicated AI compute as intact and the long-term economics of owned power and scaled GPU clusters as attractive. Recent weakness improves the entry point for a name that can still deliver significant revenue and ARR growth as capacity comes online. $INTC (Intel) โ€จStrong Data Center & AI growth, improving gross margins, and progress across both product and manufacturing. Foundry note: Based on sources we track, we identify the foundry roadmap as strong going into H1 2027. 18A is ramping with improving yields; 18A-P is in risk production with meaningful performance/power gains; and 14A remains on track for risk production in H2 2027 (with volume in 2028), supported by defect-density progress, updated PDKs, and rising customer evaluation activity. Advanced packaging (EMIB/Foveros) is also scaling to support larger AI systems, with key elements targeting 2027 production readiness. Narrowing foundry losses and a clearer path toward breakeven around 2027 reinforce the multi-year recovery case. We continue to see Intel as a recovery story that benefits from both CPU demand in AI servers and the shift toward more diverse leading-edge and packaging capacity. The risk/reward remains favorable on further pullbacks. $MU (Micron) โ€จDirect beneficiary of the HBM and high-end DRAM supercycle. Supply remains tight (management has signaled conditions could stay constrained through 2027โ€“2028), pricing power is elevated, and long-term customer agreements provide visibility. Memory is the critical bottleneck in AI systems; we expect this to remain a multi-year structural tailwind rather than a one-cycle event. $ARM (Arm Holdings) โ€” initiated โ€จWe initiated a position in Arm alongside our Intel exposure. Arm is capturing rising royalty and licensing opportunity from AI data-center CPUs (efficiency, agentic workloads, custom silicon). Pairing it with Intel gives complementary exposure across x86/foundry and Arm architecture royalties. At current levels after the broader drawdown, the risk/reward is attractive relative to the multi-year AI infrastructure buildout. Positioning & risks: We are adding selectively rather than aggressively. Key risks we are monitoring: a sustained re-acceleration in inflation that forces additional Fed hikes, further stress in highly leveraged AI infrastructure financing, and any signs of slowing hyperscaler capex or memory demand. None of these currently alter our constructive medium-term view. We will continue to buy through the noise while the fundamental AI infrastructure and memory theses remain intact.
1
1
2
1,149