EXCLUSIVE: Coatue's Public… | Sourcery Newsletter Summary

1. Key Themes

Theme 1: "Follow the Gigawatts" — Energy as the Atomic Unit of AI Investment

Coatue's investment framework has evolved from "follow the GPU" to "follow the gigawatts," treating energy capacity as the primary constraint and investable signal in AI infrastructure. The trailing one-year S&P 500 data backs this up: Semis & Hardware led all sectors at +68% and Energy came in at +18%, while every other sector posted negative relative returns.

"The gigawatt is almost the atomic unit of where the growth in AI is coming from, and it's one of the biggest shortages as well that's out there."

"The fact that it is tight, and it almost seems like it's getting tighter every week that goes by."

Memory suppliers are now signing guaranteed supply commitments through 2029 and 2030, further confirming the structural (not cyclical) nature of the shortage.

Theme 2: Sellers of Shortage vs. Buyers of Shortage — The Great Cash Flow Transfer

The single most actionable investment framework in the piece: companies that supply constrained AI infrastructure resources are dramatically outperforming those that consume them. Year-to-date, sellers of shortage (Broadcom, NVIDIA, Micron, SK Hynix, TSMC, etc.) returned ~107% on average. Buyers (Amazon, Google, Meta, Microsoft) returned just ~4%.

"When price is the main lever of your revenue growth and you have fixed costs, your operating profit actually go up multiples of what your price increases or your revenue's growing at."

The cash flow data is stark: the four hyperscalers will spend a combined $680B in 2026 capex against just $66B of total free cash flow. Meanwhile, the semi cohort generates $525B of FCF combined. Coatue's annotation: cash flow is "transferring from hyperscalers to AI infrastructure."

Theme 3: The Agentic Big Bang — Agents Launching Agents Create Exponential Token Demand

The shift from chat-based AI to autonomous agents marks a structural step-change in compute demand. Coatue labels 2025+ the "Claude Code Moment" and "Agentic Big Bang." Agents remove the human bottleneck, parallelize work 10–100x, and stretch workflows from minutes to days — from "1 analyst helping you" to "1,000 analysts working 24/7."

"You're able to launch agents, and then you're able to launch multiple pools of agents that also launch multiple agents below. And so to me, it's just the sheer power of exponential growth of the number of agents and the amount of work that's being done."

"In a year or two, we're gonna have a claw that's running hundreds of agents at all time."

Token volumes confirm the shift: annualized API tokens from OpenAI, Anthropic, and Gemini tracked near zero in early 2024, hit 16.4 trillion by late 2025, and are now on a trajectory toward 35 trillion-plus — labeled "hyper-exponential" by Coatue.

Theme 4: The Great CPU-GPU Flip — A Massively Overlooked Semiconductor Opportunity

The shift from training to agentic workloads is quietly inverting the compute architecture of AI data centers, creating a potentially enormous and underappreciated opportunity in CPUs. Previously the ratio was 1 CPU to 8–16 GPUs. It has already moved to 1 CPU to 4 GPUs — a 2x improvement — and Coatue believes it could flip entirely.

"Now the ratio is actually moving from to one CPU to four GPUs, and so it's improved kind of by 2X already, and we think it actually has a chance to flip the opposite direction, which is one GPU to four CPUs."

If the ratio flips to 1 GPU per 8 CPUs, that embeds a 16x expansion of the CPU market opportunity inside the same workloads currently driving GPU demand. The beneficiaries: Intel, AMD, and Arm.

Theme 5: Trillion-Dollar Private Giants Are Reshaping Public Market Dynamics

The historical pattern — companies becoming large after going public — has broken. Private AI companies are entering the global top 25 by valuation before issuing a single public share. This is compressing public market multiples on buyers while creating a looming IPO pipeline that could dwarf anything seen previously.

"Today you look, OpenAI's most recent round was 800 plus billion dollars. SpaceX is most recent, when they did the transaction with xAI was $1.25 trillion. Anthropic's last round was high $300 billions."

Combined ARR of OpenAI and Anthropic reached ~$55B by April 2026, having started at $3.4B in early 2024 — a trajectory that compressed 15–25 years of SaaS scaling into roughly 24 months.

"They're adding $10 billion plus or minus a month, almost $2.5 billion a week. Most of the companies in the SaaS universe don't even have $2.5 billion of ARR annually, right? They're adding that in a week."

2. Contrarian Perspectives

Contrarian 1: The AI CapEx Buildout Is Sustainable — The Bears Are Wrong on Funding

The conventional concern is that hyperscaler AI capex is unsustainable and will eventually roll over. Coatue's funding model directly refutes this. Hyperscaler EBITDA of ~$1T/year (growing at 10% annually) compounds to $6T cumulative through 2031. Add 3x debt leverage ($4T) plus sovereign funds, private credit, and Neoclouds ($2T), and the total funding capacity reaches $12T — funded primarily from operating cash flow, not financial engineering.

"They are not investing hundreds and hundreds of billions of dollars growing at these high rates into something that they think is just gonna kinda peak and trough."

The buyers are the largest, most cash-generative companies in the world. The buildout is not a bet being stretched; it is a bet being funded from strength.

Contrarian 2: A "DeepSeek Moment" Would Be Net Positive for AI — Not a Sector Killer

The consensus view treats a sudden efficiency breakthrough (like DeepSeek) as a catastrophic risk to AI infrastructure spending. Rangwalla's view is more nuanced: while it would compress sellers of shortage in the short run, it would accelerate adoption and expand the overall market in the long run. The bear case on AI efficiency gains is actually an underappreciated bull case on aggregate demand.

The biggest risk he watches is another DeepSeek moment — a breakthrough that materially reduces the power, compute, or memory required to produce the same intelligence. In the short run, such a moment compresses the sellers of shortage. In the long run, it likely accelerates adoption and expands the pie.

This is especially important because Jevons' Paradox has historically held across every major technology efficiency cycle: cheaper compute has always led to more total compute consumed, not less.

Contrarian 3: The Mag 7 Is Not a Monolith — Internal Dispersion Is Extreme and Growing

The market still largely treats the Mag 7 as a unified AI beneficiary basket. Coatue's data shows this framing is dangerously wrong. Since the October 2025 market peak, Google is up ~+40%, Amazon up ~+20%, and at least one Mag 7 name is down more than 20%.

"Even within the Mag 7, the reordering of the pecking order is happening quickly." — Philippe Laffont

The differentiating factor: Google (via TPUs) and Amazon (via Trainium) own silicon as well as deploy it, giving them a partial hedge on the seller-of-shortage dynamic. Pure buyers with no silicon ownership face multiple compression and rising input costs simultaneously.

"I actually think Amazon and Google are in an unusual camp, and those are the two companies that we really like because they are also a bit of a hybrid in that Google has TPUs that they sell, and Amazon has talked about selling Trainium."

3. Companies Identified

NVIDIA

Leading GPU manufacturer; original "follow the GPU" thesis
Why mentioned: Coatue's foundational early bet; now part of the seller-of-shortage cohort generating $184B in FCF; beat Q2 2023 earnings estimates by ~30%, catalyzing the AI investment cycle
Quote: "We bet on Nvidia early, but that's usually not enough."

Anthropic

AI frontier model company (Claude)
Why mentioned: Valued at ~$300B+ in latest round; combined with OpenAI driving ~$55B ARR by April 2026, adding ~$2.5B/week; "Claude Code Moment" named as the catalyst for the Agentic Big Bang
Quote: "Anthropic's last round was high $300 billions."

OpenAI

Frontier AI model company (ChatGPT, GPT series)
Why mentioned: Valued at $800B+; combined ARR with Anthropic hit ~$55B by April 2026; ChatGPT named as the moment that "changed the interface" in 2022
Quote: "OpenAI's most recent round was 800 plus billion dollars."

SpaceX

Aerospace and satellite company
Why mentioned: Valued at $1.25T (post-xAI transaction); cited as one of three private companies now inside the global top 25 by valuation
Quote: "SpaceX is most recent, when they did the transaction with xAI was $1.25 trillion."

Micron

Memory semiconductor manufacturer
Why mentioned: Prime example of seller-of-shortage operating leverage; operating margin expanded from 16% five-year average to 69% today (+4x); signing supply commitments through 2029–2030
Quote: "Micron's operating margin has gone from a 16% five-year average to 69% today (+4x)."

Intel

CPU manufacturer
Why mentioned: Coatue's CPU-GPU flip thesis makes Intel a key beneficiary; highlighted specifically as a constructive position with new leadership
Quote: "Some of the best thesis are the simple thesis, and you don't wanna overthink it."

TSMC

Semiconductor foundry
Why mentioned: Listed as a seller-of-shortage generating significant FCF ($46B); part of the ~$525B semi cohort FCF figure

SK Hynix

Memory semiconductor manufacturer
Why mentioned: Seller-of-shortage generating $110B in FCF; part of the tug-of-war framework on the supply side

Seagate

Data storage company
Why mentioned: Operating margin expanded from 17% to 38% (+2x) as a seller-of-shortage case study

Broadcom

Semiconductor and infrastructure software company
Why mentioned: Named as a top-performing seller of shortage in the ~107% YTD return cohort

GE Vernova

Energy infrastructure / power generation company
Why mentioned: Listed on the seller-of-shortage side of the tug-of-war; energy as a constrained input to AI buildout
Quote (context): Part of the "follow the gigawatts" thesis; energy sector returned +18% trailing year, second only to Semis & Hardware

Google (Alphabet)

Hyperscaler / search / cloud
Why mentioned: One of two "hybrid" companies Coatue favors among buyers because it also sells TPUs; Google up ~+40% since Oct 2025 peak — best Mag 7 performer
Quote: "Google has TPUs that they sell."

Amazon

Hyperscaler / e-commerce / cloud
Why mentioned: Second "hybrid" company Coatue favors; developing and selling Trainium chips; up ~+20% since Oct 2025 peak
Quote: "Amazon has talked about selling Trainium."

Meta

Social media / AI infrastructure buyer
Why mentioned: Forward P/E compressed from 23x to 17x since Jan 2025; spending $138B in 2026 capex with only $5B FCF; named as pure buyer of shortage facing multiple compression

Microsoft

Cloud / enterprise software / AI buyer
Why mentioned: Forward P/E compressed from 30x to 22x since Jan 2025; spending $157B in capex; named as pure buyer with no silicon offset

Samsung

Memory / semiconductor manufacturer
Why mentioned: Seller-of-shortage generating $132B in FCF; part of the semi cohort

4. People Identified

Jaimin Rangwalla

Title: Chief Investment Officer of Public Investments, Coatue Management
Why mentioned: Primary interview subject; author of the Spring 2026 Public Markets Update; 20-year Wall Street career; joined Coatue in 2007 when it managed $700M with 12 people
Quote: "I feel like we're, as an organization, are at our second inflection. Maybe, probably not second, like fifth inflection. But we're at another inflection where I think we're really on this journey to really just break out."

Philippe Laffont

Title: Founder/CIO, Coatue Management
Why mentioned: Quoted directly on the Mag 7 reordering dynamic during the investor call
Quote: "Even within the Mag 7, the reordering of the pecking order is happening quickly."

Michael Barton

Title: Sector Head, Coatue Management
Why mentioned: Pictured in the article at Coatue's Tech Museum; identified as a key member of the coverage team

5. Operating Insights

Insight 1: Restructure Your Coverage and Team Around the AI Supply Chain, Not Traditional Sector Buckets

Coatue has fundamentally reorganized its analyst coverage to map onto the AI supply chain itself — compute, energy, memory, networking, model layer, applications — rather than traditional sector categories (software, hardware, internet). For operators building research, sales, or product teams, the same logic applies: org design should follow where value is being created and destroyed, not legacy industry classifications.

"We're almost thinking what slice of the AI supply chain, and then who are the people we're gonna have covering those slices."

Insight 2: Use the Seller vs. Buyer of Shortage Framework to Diagnose Your Own Business's Position

Any business involved in the AI ecosystem should ask: are we a seller of something that is scarce and price-inelastic, or a buyer of something that is getting more expensive? The operating leverage math is extreme in favor of sellers: a 2x price increase on a fixed-cost base can produce a 3x jump in operating profit; a 4x increase can produce an 8x jump. Understanding which side of the tug-of-war your business sits on is the first step to either defending margin or reorienting strategy.

"When price is the main lever of your revenue growth and you have fixed costs, your operating profit actually go up multiples of what your price increases or your revenue's growing at."

Insight 3: Anchor to Fundamentals During Volatility — Use It as a Signal, Not a Reason to Sell

The MSCI USA Momentum Index has logged 70 days YTD (annualized) with moves of +/–1.5% or more — the most since COVID. Rangwalla's discipline in high-volatility environments is to return to fundamental analysis rather than react to price action, and to use hedges, puts, and shorts as volatility management tools rather than abandoning core positions.

"The volatility we're facing now is, wow, I'm still really bullish. AI is still doing great things, but wow, some days, some of our stocks are just down 5% or 10% on for no reason."

"At the end of the day, fundamentals do matter more so than sentiment, and I think the fundamentals are really strong."

6. Overlooked Insights

Overlooked Insight 1: The $6T+ TAM Includes Physical AI — And Coatue Thinks Even That May Be Too Conservative

The article's TAM discussion focuses heavily on digital AI ($4T+), but Coatue explicitly adds a physical AI layer — humanoids, industrial automation, and autonomous vehicles — to reach $6T+. More importantly, Coatue's own annotation on the slide reads: "Could be much bigger!" This suggests the firm's published TAM estimate is itself a floor, not a ceiling, and that physical AI is an underweighted component of most investor AI frameworks.

The $2T coding/developer TAM alone assumes capturing only 40% of the $5T global engineering and IT payroll — meaning even that single category has significant room to expand if AI penetration rates prove higher than modeled.

Overlooked Insight 2: OpenClaw Is Emerging as One of the Fastest-Scaling Repositories in GitHub History

While Anthropic's Claude is widely discussed as a frontier model, the specific developer tooling layer built around it — OpenClaw — accumulated roughly 190,000 GitHub stars in its first ~4 months, making it one of the fastest-scaling repositories on GitHub ever. For context, React took ~100 months to reach ~215,000 stars. This trajectory suggests that the orchestration and tooling layer around agents may be a faster-moving and less-appreciated investment surface than the models themselves.

"One of the fastest-scaling repositories on GitHub... In a year or two, we're gonna have a claw that's running hundreds of agents at all time."