10 Year Hedge Fund Analyst | Private Company CFO | Sharing idiosyncratic stock ideas with accelerating fundamentals | NFA DYOD
New York, USA
Joined July 2023
- Tweets2K
- Following2.1K
- Followers14.9K
- Likes7.4K
Excellent summary on the state of AI within the enterprise
Spoke to an AI leader recently.
The biggest thing they’re wrestling with: what do they do with the AI laggards in their company?
Like most companies, they’ve found that a Super Pareto Principle is true within their org. 10% of their employees are responsible for 90% of the uplift in productivity/token usage from software engineering to non-technical knowledge work.
They’ve spent a lot of time & resources on AI enablement from LLM foundations to function-specific trainings and skill building. And yet, the power law of productivity increases hasn’t budged.
So now they’re left with a few sobering thoughts and hard questions:
1) 90% of people are overpaid and 10% are underpaid
2) We’d be better off allocating tokens proportionally to the power law, but how do we do that without killing culture?
3) Significant layoffs are likely warranted and wouldn’t impact overall output, but cultural backlash is a big concern?
4) How do we hire more power law people as a non technology company?
5) Is all of our efforts on training and enablement ACTUALLY worth it?
6) How do we not experience a brain drain where the top 10% all leave because they want to be surrounded by other AI-native 10x’ers?
There's a new version of this post
I think the bottom in neoclouds is in:
Two of the best thinkers @GavinSBaker and Elon on the topic have made the thesis very clear:
1. Gavin — token prices can drop while revenue per GPU up
2. Elon — paraphrasing here but “even if you pay $38/gpu hour the incremental productivity (ie ROI) is so high relative to human labor that SI/AI should be used wherever possible”….. this is clear from Anthropic’s S-1 where revenue is exploding relative to compute
So you have more ppl realizing the huge arb between agents and people. You have more demand for GPUs, GPU pricing goes up….
Paradoxically the market could be punishing $crwv for having 5 year contracts locked in solid rates (but unfavorable to where pricing is today) during 2024/2025…. Same dynamic happened with banks locked into low mortgage rates
This creates a lot of opportunity for neoclouds like $brun who are executing well but have less of their fleet tied to sub optimally priced deals
Pretty bullish $crwv $nbis $brun
Mega Cap Tech has nearly doubled revenue last five years while keeping headcount flat. The ROIC from AI is accelerating
$meta $msft $googl $amzn
@Jason @chamath @DavidSacks @friedberg @GavinSBaker @altcap
$msft
The reason Wall St. has so much trouble understanding the ROI of AI (which is so blatantly obvious) is because the workflows of equity analysts don’t align. It’s not a knock, but it is the reality.
If you’re mostly in Bloomberg/FactSet, reading research, using Claude to generate research and financial models, your lived experience on the impact of AI will be very different than how a high agency corporate executive will see it.
Most equity analysts have never interacted with an ERP, CRM, Supply Chain, HCM software, tried getting cross-functional teams (R&D, legal, compliance, purchasing, accounting, etc) on same page across org. It takes/wastes a LOT of time. The impact that AI can have on improving accuracy, speed, while also reducing costs is mind boggling.
This is a good example: 45x engineers. Engineers now doing 45x more per engineer. Assuming 300k comp, that is $13m in more output. ROI is obvious.
My take from the Anthropic S1 today is that they are no longer reading all their code. Even for the engineers directing AIs to create it, they're not reading every line. This should be obvious to everyone by now, but I suspect some engineers are still in denial.
Two direct quotes lead me to this: "over 90% of the contributions to our codebase in September 2026 were written by Claude." and "the average daily volume of code contributed to our codebase per engineer in the three months ended September 30, 2026 was over 45 times the average daily volume in the year ended December 31, 2024."
For a company with some of the best engineers on the planet, they're creating 45x the volume from before. I guarantee you they can't actually read all of that.
You have to pull the humans up to a higher level of abstraction to get that kind of velocity.
“No ROI on AI”…..
My guy, $MSFT is cutting 27% of its employees and growing 18% YoY… they expect to continue that rate of growth with 27% fewer employees…..
The ROI is in fact, massive
Incremental return over incremental capex. The concept blows peoples minds for some reason.
With respect to AI capex, 5% price increases with 2% headcount reduction produces a 10% ROIC for the hyper scalers. This is due to their incredibly high incremental margins and pricing power.
Aggressive but no bubble.
Thoughts @mjmauboussin @ContrarianCurse @altcap @bgurley @chamath @friedberg @BillAckman
As I’ve said, AI is way cheaper than humans, way smarter than humans, way more persistent…. The autonomous enterprise is coming. Margins of companies on the forefront will explode higher… the cuts are just beginning, they will be deep at companies with an AI forward approach (ie companies that will win)
Diffusion is happening. Big orgs move slower because of internal politics, but it’s happening across industries. The hallowing out will be dramatic…..
This will cause a lot of issues… rich getting richer, unemployment up, rates down (less inflationary pressure from labor market), stocks and real estate go higher, virtuous cycle up. Many will get left behind. Socialism will rise
Strangely, it seems the smartest people in the world who invented AI actually have a better understanding of unit economics and business than the financial advisors on CNBC telling you that they're cautious on AI and could be bubbly
xitter.cf/BitcoinAIGuy/status/21…
Asymmetric Bets retweeted
Anthropic "is burning cash like crazy" was one big lie, spread by the Ed Zitrons of the world.
For the first 6 months of 2026, Anthropic generated positive free cash flow of $3.4 billion ($3.9 billion of cash flow from operations less $482 million of capex).
And stock-based compensation was only $2.2 billion.. So even FCF less SBC was positive.
Remember in June/July all of the idiots on the timeline saying $meta doesn’t have demand for its chips and will be cutting capex? I do. Well today they’re raising $35B to buy more chips because they’re short….The misdirection in AI has been insane.
$META announced plans to raise approximately $35 billion through two offerings:
> $22.5B in Class A common stock through forward sale agreements.
> $12.5B in depositary shares representing mandatory convertible preferred stock.
Interestingly, NVIDIA, AMD and Broadcom have expressed nonbinding interest in purchasing up to $5 billion of Meta shares combined.
Meta is also planning a potential $60 billion offering program for its new nonvoting Class C shares starting in Q1 2027.
Together, the plans could provide approximately $95 billion in equity financing capacity as Meta prepares to spend $135B to $145B on capital expenditures in 2026.
Cannot disagree with this more. If you can’t understand that AI is 25x cheaper than employees and is infinitely persistent, you simply aren’t using the coding tools.
Autonomous enterprise is going to be a term that’s used a lot more in the coming 12-18 months. Then the layoffs will accelerate. This is obvious to anyone using the tools…. Moreover rising prices are also obvious over the next few years
Asymmetric Bets retweeted
The AI bears are extremely active today, particularly in relation to credit markets. Here are some thoughts to keep your timeline balanced.
✅ Nebius priced a $3.5B Citi facility at 6.7% on 22 October.
✅ Cipher priced $850M of 0% convertible notes on 14 October at a 35% conversion premium. Investors took no interest at all for exposure to the upside.
✅ Lenders are pricing the customer. Debt backed by hyperscaler leases (AWS, Google, Oracle) trades at about 6.5 to 8.5%, while CoreWeave's unsecured bonds trade above 15%. CoreWeave's own Meta-backed GPU loan priced at SOFR + 2.25%.
✅ The long trend is still cheaper money. Hyperscaler-backed deals are pricing in the mid-6% range this year, compared with 9 to 10%+ in 2025. CoreWeave's cost of capital went from about 15% in 2023 to about 6% in 2026.
✅ GPU rents are up this year on every generation: B300 +56%, B200 +40%, H100 +30%, A100 +14%.
✅ Bitcoin miners have signed 40+ AI hosting deals since mid-2024, covering about 10.9 GW and roughly $188B of contracted revenue.
Stealing companies is a serious crime
$fnma $fmcc
@pulte @SecScottBessent @realDonaldTrump @howardlutnick @DonaldJTrumpJr @BillAckman
Asymmetric Bets retweeted
The way to think about this is:
How many super smart and dedicated people do you want to hire? And they keep getting better.
Zooming out, GPUs are still renting ahead of a standard 6-year depreciation schedule
Rental price vs. schedule:
- H200: +143%
- B200: +111%
- RTX PRO 6000: +73%
- H100: +54%
- A100: +9%
Most of the gap opened up after April, when token demand took off
Charts of the Week: a16z.news/p/charts-of-the-we…
The viral Apollo charts claiming "AI spending is just circular VC money" are a masterclass in financial data misdirection. Here is why the narrative collapses the moment you look under the hood:
1. The Ultimate Selection Bias (The Ramp Dataset)
Apollo’s data comes from Ramp corporate card swipes. Who charges compute on a corporate credit card? Seed and Series A/B tech startups. Who does NOT? JPMorgan, Walmart, Apple, or the DoD. Real Fortune 500 enterprise spend flows through multi-million-dollar negotiated master service agreements via AWS Bedrock, Azure, or Google Cloud Marketplace—not an expense card. Running corporate card data to measure global enterprise AI adoption is like surveying a Y Combinator cafeteria to track global dietary habits.
2. A Basic Math Trick: Power Laws Exist Everywhere
Apollo highlights that the top 10% of firms account for 95.6% of OpenAI and Anthropic spend. Sounds terrifying until you glance two lines down on their own table: the top 10% of firms also account for 95.7% of standard cloud computing and 91.9% of paid ads. Corporate spend is logarithmic. A 5,000-person company naturally spends 500x more on infrastructure and CRM than a 10-person shop. Framing standard enterprise power-law distributions as an "AI-specific bubble" is intellectually dishonest.
3. Intentional Wordplay
"80.4% of the top 1% of AI spenders are VC-backed" does NOT mean 80% of total industry revenue comes from VC startups. It means within a startup card dataset, the heaviest spenders are venture-backed labs training models. Meanwhile, Apollo casually ignores the bar directly beside it: the median AI spender is 86.9% non-VC—proving broad, organic adoption across traditional businesses.
4. Follow the Incentives
Apollo is a massive private credit and alternative debt titan. Their house macro view thrives on convincing institutional allocators that tech equity multiples and venture ecosystems are unstable, steering dry powder directly into senior secured private credit and high-yield asset-backed lending.
FUD gets clicks, but using corporate card data from startups to argue that the entire AI economy is fake isn't serious macro analysis—it’s narrative engineering
@altcap @GavinSBaker @chamath @friedberg @Jason @DavidSacks @elonmusk @BillAckman
The idea that the ten year yield moving from 4.5% to 5% will kill the AI build out is peak Wall St. brain rot.
A 50bps swing does not matter when hyperscalers and labs are fighting a generational land grab where having compute 6 months early is worth $20B in high margin revenue.
50bps on a $50B 1GW data center is $250M. $250M more in interest vs. $20B in high margin revenue. The $250M of interest is 1.25% of revenue. Most data centers are also not 1GW....
The opportunity to buy infrastructure stocks here is massive $crwv $nbis $ORCL $BRUN
@GavinSBaker
The “AI is just VCs trading money in a circle” narrative collapses once you look at actual adoption.
Revenue growth isn’t dangerously concentrated in a few tech giants - $META cutting 30k internal seats on Claude barely moved Anthropic’s topline by 1%. The bulk of ARR is thousands of real enterprises and SMBs paying for reliable API tokens.
Open source won’t kill frontier labs for the same reason Linux didn’t kill AWS: “free software” isn’t free once you have to pay $25k/month on GPU clusters and hire an infrastructure team to keep it alive at 3am. That aint free.