Lizardinho (@Lizardinho13) header
Lizardinho (@Lizardinho13) profile picture

Buyer of highs, seller of lows.

Cayman Islands
Joined February 2022
Photo by Lizardinho (@Lizardinho13)Photo by Lizardinho (@Lizardinho13)Photo by Lizardinho (@Lizardinho13)Photo by Lizardinho (@Lizardinho13)Photo by Lizardinho (@Lizardinho13)Photo by Lizardinho (@Lizardinho13)Photo by Lizardinho (@Lizardinho13)Photo by Lizardinho (@Lizardinho13)Photo by Lizardinho (@Lizardinho13)Photo by Lizardinho (@Lizardinho13)
Lizardinho retweeted
The craziest part about neoclouds ($NBIS , $IREN, $CRWV) crashing is the timing. Stocks crash. That's normal and healthy market behavior. But what's remarkable here is that neocloud are collapsing precisely as their underlying market is experiencing explosive growth at an unprecedented pace. Part of the reason for this crash is a distorted supply/demand view. Neocloud and hyperscaler earnings gave us a glimpse into the supply side of the equation. That's what the street is focused on now. But the Anthropic's S-1 gave us the first look into the demand side, and that's what many investors are completely missing. In Q2'26, $ANT was spending $5.6B on compute to support $11.6B in revenue. By Q3, revenue had climbed to $17B, and compute costs grew to $9.9B. That's a 77% increase in compute costs in a SINGLE quarter, outpacing even revenue growth. Understanding the source of growth makes this even more interesting. Much of this growth is coming from enterprise customers, which usually have a significant step up in usage after joining. So in Q2, Anthropic had 6,000 customers spending more than $100K annually. By Q3, that number had nearly doubled to 11,900. By the end of Q4, it's expected to approach 18,000. That's potentially a tripling of large enterprise customers in just six months. By my estimates, these additional customers alone will require at least $3B in incremental compute spending, bringing Anthropic's FY26 compute bill to roughly $31B. By 2027 compute spending will be AT LEAST $55B, assuming revenue growth slows dramatically (after three consecutive years of roughly 10x growth). Very likely, compute will be MUCH higher because there is no evidence to show that growth is actually slowing this much. So Anthropic's compute needs are set to nearly double AGAIN next year, while existing capacity is already stretched to its limits. Where is all this additional compute going to come from? AI labs cannot grow without compute. They can sign all the customers they want, but without the infrastructure to serve them, that revenue simply cannot materialize. This is precisely why $ANT is reportedly willing to pay $SPCX as much as $50M per MW of compute capacity, roughly 3x the prevailing one-year spot rate. Because the alternative is far worse: They either secure the capacity, or they stop growing. This is the disconnect I find so extraordinary. It's fine for stock to drop in price. It's just odd for that to happen in the middle of this historic demand explosion. Please comment and share if you enjoyed this read.
Few people realize just how fast Anthropic ($ANT) is growing. By my estimates, the company is currently facing a $20B shortfall in secured compute capacity for 2027 alone. To put demand into perspective, the number of customers spending over $100K annually nearly doubled from 6,000 in Q2'26 to 11,900 in Q3'26. These customers generate an average of roughly $2.5M in annual revenue. The current compute footprint simply cannot support the $ANT growth without additional capacity. I'm assuming that @OpenAI it is growing at a similar pace and is facing similar compute constraints. This is why I'm so bullish on neoclouds like $NBIS. The demand for compute is accelerating and someone needs to build and deliver all that additional capacity. This is simply math.
43
82
8
691
99,499
Risk off signals have been flashing for a while now (thanks @themarketradar) and reading this convinced me that the bottom on AI is not in yet. Gonna take some time.
$AMD一天跌了10%,比$NVDA跌得还狠,这是今天最该看懂的地方。 周四黄仁勋离职那天,资金的第一反应是从$NVDA里撤出来,转去$AMD、$AVGO,赌英伟达一乱,订单会流到对手那边。结果今天这笔交易被整个平掉了:盘中$AMD跌了约10%,$AVGO跌了5%左右,$NVDA反而只跌了2%到3%。挪过去的仓位一天就撤回来,说明市场根本不是在押谁抢谁的份额,而是在给整条AI链一起打折。 打折的理由,今天也摆上台面了。 黄仁勋在的时候,$NVDA一直给下游的新云公司做融资、做担保,用自己的钱帮客户买自己的卡。人一走,市场最先担心的就是这套循环还转不转得下去。所以今天跌得最狠的是$CRWV、$NBIS这类新云,盘中都跌了9%左右。英国的Nscale改了招股书,前九个月净亏损扩大到17亿美元。$MSFT也传出在重新审视跟新云的合同,打算少用几家。光模块$LITE、$COHR,服务器$DELL、$HPE,跌了6%到11%不等。 钱从哪个口子往外流,看跌幅排序就知道:最靠融资活着的先跌,靠$NVDA订单吃饭的跟着跌,现金流最厚的巨头跌得最少。 利率也在帮倒忙。10年期美债周四重新站上5%,今天盘中涨到5.1%以上,30年期在5.4%附近。PPI其实比预期温和,债市却不买账。估值高、资本开支重的科技股对这个数字最敏感。纳指100盘中跌了约1.7%,道指基本没动,必需消费反而在涨。 再就是下周的$ANT。 下周三定价,周四上市,8.25亿股,每股70到78美元,最多能募600多亿美元。这么大的认购量,基金得提前腾出现金,最顺手的办法就是卖掉手里涨过、流动性好的AI股。$NVDA自己还承诺在配售里认购100亿美元,现在黄仁勋走了,这笔钱会不会变,市场也在猜。 三件事叠在一起,今天跌成这样一点都不奇怪。 那为什么说这还算不上崩盘? 因为跌的只是AI这一条线,大盘没有一起塌。标普盘中跌不到1%,道指持平。资金是在科技股内部换位置、往防守板块躲,不是什么都卖。11/5那天费半单日跌12%的那种恐慌,今天没有出现。 位置也要算清楚。费半到周四收盘,比上周五高了12%左右,前面反弹得太快,今天有一部分是在吐这段反弹。 接下来就看两件事。一是11/17 $NVDA财报,新管理层怎么讲订单,怎么讲对客户的融资。二是$ANT的定价和首日表现。$ANT发得顺,说明一级市场对AI资产还肯给价,被抽走的钱有机会回流。要是定价压在区间下沿甚至破发,这轮下跌大概率要回去测11/5的低点。 $AMD这种一天跌10%的刀,这周我不去接,也不赌$NVDA的财报夜。反弹先减仓,跌到位置再分批买,现金留着过完下周。 公司还是那些公司,只是现在没人愿意在$ANT上市前多扛一点仓位。
1
15
Lizardinho retweeted
Anthropicの最新決算、言葉が出ないほど酷いなこれ、、 擁護する余地を探そうとしたけど見当たらない 計算コストの圧迫で調整後利益率がこのフェーズですら悪化するって、 中国LLMとの価格競争起きたらどうなるんだ😓 しかも調整後営業利益のQoQ成長率は大幅鈍化してるし
12
153
28
998
340,555
To specify, the issues in tech are: - OAI fund raise undershot goal - Antropic Q4 growth (so far) 15-25% vs 50% Q3 - Kioxia earnings - Meta 40% chips idle (Alphasense report) - Early adopter on premise open weight risk (e.g. Revolut/Bending Spoon) teapawt.org/fkronawitter1/status/2…
We got another bounce. I am not sure about its longevity this time as there are still many issues in Tech teapawt.org/fkronawitter1/status/2…
17
19
2
258
83,002
Lizardinho retweeted
The next time someone tells you that markets are rational remember that in 2026 pretty much every company simultaneously announced that they were going to borrow a quadrillion dollars to build 420 data centers in six months with chips they didn’t have powered by electricity they couldn’t get from turbines that weren’t built sent through transmission lines that didn’t exist and everybody was like “Cool, cool. Sounds good. Solid plan.”
102
1,135
60
9,121
993,064
Most people voted already in key states. Seems too late for Trump to capitulate now. Short everything.
19
AI related shorts closed, thanks for playing. QQQ sitting at 200D EMA. Bidding to play the bounce. Would not be surprised if Cheetoman pulls a rabbit out of the hat during the weekend
13
Google launches Gemini 4 Carbon frontier model, priced at $2/m input, $10/m output tokens $GOOGL
2
4
28
46,148
Lizardinho retweeted
He said WHAT?
$NBIS CEO Arkady Volozh Says Debt Markets Closed for Neoclouds Since Last Week "We have a pretty healthy balance sheet. We always thought about being more conservative here. We used all kinds of instruments. From last week, probably some of the capital sources are temporarily closed, like the asset-backed or debt market. But we're working now with customer advances and several other instruments,which will probably provide us enough finance to build this gigawatt at least."
2
5
36
4,716
Lizardinho retweeted
Goldman forecasted AI agent monthly demand of 120 quadrillion tokens by 2030. If every Grace Blackwell 300 served DeepSeek v4.1 Flash, we'd be capable of generating 13,200 quadrillion tokens a month already. 100x demand. This is the issue I think @elonmusk is raising.
Replying to @Yuchenj_UW
I think we’re going to see a lot of stranded compute in partially complete data centers
9
10
4
99
15,052
Lizardinho retweeted
Replying to @value_invest12
I genuinely wonder how this can make any sense?
2
9
1,585
READ: if everyone used DeepSeek’s ultra efficient v4.1 flash model there would be such a gargantuan compute glut already that we could just start canceling a bunch of the future compute commitments. The reality: this isn’t going to happen overnight but this is the direction we are going and we haven’t even stopped the logarithmic efficiency improvements. Yes token consumption will grow exponentially but there are so many supply exponentials that will completely overwhelm demand. COMPUTE GLUT.
Western team have independently verified that DeepSeek's performance numbers of 38$ dollar revenue per GPU at DeepSeek's rocket bottom prices is true. This is a momemnt when you update your priors. At this rate with just 1 GW, you can make a revenue of 150B per year. Even with lot of slack it would be 100B per year. If you use half the compute to come online by 2027 end, you will have 25GW for inference with potentially 2.5T revenue at DeepSeek rates and 10T at USA lab rates (world GDP is 125 T, 10T is 8% of it) You answer for your-self are we overbuilding!!!
10
11
1
113
10,695
Lizardinho retweeted
the OpenAI discounting is ~serious~ and is have an impact
24
17
24
512
103,110
New model every day, cheaper tokens every day. No wonder nobody wants CRWV credit. They won't be able to pay their bills
Kimi K3.1 10x cheaper than Opus 5.5 at Sponge Bob 🍔 Teams of 5 @Kimi_Moonshot and Opus 5.5 agents got the same prompt and Kimi finished in 26 minutes while Opus took 2 hours and ended up costing much more Run both models via API -> atomic.chat
56
Ken Griffin of Citadel questions the economics of the AI infrastructure boom: ›What happens when GPU supply exceeds demand & pricing goes into free fall? ›Are today's semiconductor margins sustainable? Are investors overestimating the useful life of GPUs? “Look, I think this market is actually experiencing people doing research. For the couple of years, if you were on the thematic program of we're gonna spend hundreds of billions or a trillion dollars in data centers, we're gonna build compute at a scale never seen before, which companies in that story are choke points in supply chains that will have abnormally high margins? And who will be the big winners in this transformation? Will it be the Microsofts & AWSs who will be the gatekeepers to the foundational models? Like, who's going to earn the economic rents? But in some sense, like, 'I'm just gonna buy them all. I'm gonna buy the hyperscalers, I'm gonna buy the chip, I'm gonna buy businesses that make things I've never even heard of 5 years ago in Korea and Taiwan & in Japan.' And now we're reaching a point where people are becoming more discerning. We do realize that we are seeing extremely high margins in DRAM & MLCCs in areas where there are real supply constraints, but people are now starting to look out 2 years, 3 years. How long is this going to last? And in particular, I don't wanna be doom & gloom, but GPUs, like airlines, have a very low marginal cost to run. The incremental seat on an airline that's not full, if you can sell that ticket for $20, you're making money in that seat. And that's one of the reasons that the airline industry is such a nightmare, is because the moment you have overcapacity, pricing goes into free fall. We saw this actually with the telecom build-out back in 2000. We built more fiber than we had demand for, & the price of telecommunications just went straight down. If we reach a point where the provisioning of GPUs exceeds demand for GPUs, marginal pricing will actually go into free fall. And that really upends all the investment assumptions that people are working with today. That's issue number one. And issue number 2 is: I think people are taking a step back & saying, 'What is the useful life of a GPU?' Right here, right now, there's people that are earning extraordinary pricing on GPUs that are 5 or 6 years old. And so there's a growing set of concerns that I can't give the same capitalization rate or the same revenue models to year 7, 8, 9 of a chip in the future that maybe I was using 12 months ago. So I think we're really grappling with just how far will this demand carry us in the marketplace.”
57
63
40
575
202,151
Don’t overthink it. Hyperscaler and AI CapEx stocks are dumping because the assumption that there will be unlimited cheap debt to fund the AI buildout is proving false. Only two ways around it: 1) Cut the CapEx 2) Fund it with equity And both result in lower EPS. Welcome to your first commodity cycle, Muse bros.
53
91
20
1,415
124,864
Replying to @stoicsavage
faded?
1
2
203
SpaceX lists ~Jun 12 at a ~$1.75–2T valuation — roughly 100x sales. $SPCX Every take is arguing whether that number is "fair." Wrong debate. The edge is in the forced-flow calendar nobody's modeling 🧵
1
7
2
23
6,855
I think that was the blowoff top for $CAR
1
192