NVIDIA Corporation NVDA
AI accelerators
AI
Research as ofOur investment view
PredictionThrough 2028: AI use and revenue will keep growing, but installed compute and power commitments will outrun profitable demand for many projects. The strongest distribution platforms and low-cost infrastructure operators will earn returns; weaker projects will be repriced, delayed, or consolidated. This is a split market, not a prediction that AI demand disappears.
What would confirm it: Rising paid usage together with uneven data-center utilization, renegotiated capacity commitments, and widening cash-return gaps between suppliers and operators.
What changes our view: Broadly high utilization and durable positive cash returns across new capacity, with disclosed AI revenue covering depreciation, power, and financing costs.
AI-capex stress and critical inputs sharpen market bifurcation
–
This week produced three developments that sharpen the split between firms that earn from AI deployments and those that consume capital: the U.S. added helium to its Critical Material List, Alphabet reported a severe market reaction to earnings tied to AI features and large AI spending commitments, and AMD announced Helios racks live at Meta alongside a $1 billion equity investment in Meta's offering.
U.S. adds helium to Critical Material List
The Department of Energy added helium to its Critical Material List and highlighted helium as non-substitutable for semiconductor manufacturing, leak testing natural gas infrastructure, quantum computing and other scientific uses. Because helium enables advanced semiconductor manufacturing and data center supply chains, constrained helium availability could raise input cost and schedule risk for chip and data-center projects, while an expanded supply or substitutes would blunt that pressure. [1]
Alphabet stock plunges after AI-linked results and multiyear spending disclosure
Alphabet shares fell sharply after quarterly results showed advertising shortfalls tied to an "AI Mode" search change, even as the company highlighted over 1 billion users of its Gemini app and disclosed over $1 trillion in multi-year AI infrastructure spending commitments; the share move erased over $550 billion in market value and fell as much as 15.8%. The scale of announced AI commitments intensifies the question of when AI capex converts into recurring revenue. [2]
AMD Helios systems go live at Meta; AMD invests in Meta equity offering
Advanced Micro Devices reported that Helios rack-scale systems are installed and testing at Meta, and that AMD will invest $1 billion in Meta's equity offering; the stock reaction included a 5.2% climb. The Meta deal and testing position AMD to capture direct hardware demand, while Meta's increased fundraising target of 97.5 billion implies large customer-side financing activity to support its 2026 capital program. [3]
Our investment read
These events reinforce our prior view that AI spending is bifurcating winners that can monetize AI and hardware providers that depend on a few large customers. If helium supply tightness persists, data-center builders and chip fabs could see higher costs and schedule risk, which could pressure companies that operate or build capacity such as Vertiv Holdings Co, Eaton Corporation PLC and Constellation Energy Corp if they face higher input costs or delayed projects. Conversely, GPU and accelerator suppliers such as NVIDIA Corporation could benefit if customers prioritize faster accelerators over bespoke in‑house alternatives; Microsoft Corporation, Amazon.com Inc and Alphabet Inc Class C could see their cloud and AI compute economics change if they either reduce external spend or scale own silicon. A conditional countercase that would weaken our thesis is observable: if Alphabet posts sequential recovery in advertising revenues tied to Gemini usage and demonstrates measurable reduction or rephasing of its over $1 trillion in commitments, then the argument that AI capex will widen cash shortfalls would be undermined.
What matters next
Monitor whether Alphabet reports a reversal in advertising revenue trends or updates the pacing of its multi-year AI commitments; track Meta's actual proceeds and deployment from its equity offering and whether AMD reports scaled Helios deployments beyond initial testing; watch DOE or industry announcements that increase helium supply or indicate substitutes, which would ease input-cost pressure.
Latest developments
Loading briefing…
Featured company research
Holdings …
Microsoft Corporation MSFT
Cloud and AI distribution
Amazon.com Inc AMZN
Cloud and AI compute
Alphabet Inc Class C GOOG
Cloud and AI compute
Vertiv Holdings Co VRT
Data-center power and cooling
Eaton Corporation PLC ETN
Electrical infrastructure
Constellation Energy Corp CEG
Power supply
Investment context
AI is an investment chain: users and businesses pay for applications; model companies buy compute; cloud providers build capacity; chip, networking, cooling, and power suppliers sell the equipment. The central question is whether revenue and useful output will grow enough to earn a return on that capacity. A contract announced upstream is not the same as a profitable, fully utilized data center downstream.
What is observable
| Evidence | Dated fact | What it cannot prove |
|---|---|---|
| Electricity demand | A 2024 DOE/Lawrence Berkeley study estimated data centers used 176 TWh, or 4.4% of U.S. electricity, in 2023 and projected 325–580 TWh in 2028. | This is a scenario range, not a measured 2026 load or a guarantee every planned campus will connect to the grid. |
| Private AI funding | OpenAI said it closed a $122 billion financing round in March 2026 and described broad cloud and chip partnerships. Anthropic announced a $65 billion financing in May. | Funding provides capacity to spend; it does not establish audited operating cash flow or future public-market returns. |
| Commercial traction | OpenAI's March statement said it was generating $2 billion in revenue per month and more than 40% of revenue came from enterprise. | These are company-reported run-rate and mix claims, not a complete public filing with margins, retention, or cash-flow notes. |
The bottleneck can move. A chip shortage can shift toward power and interconnection; abundant capacity can shift the question toward utilization and pricing. Model efficiency is double-edged: lower cost per task could improve margins and expand demand, but it could also make earlier hardware forecasts too high. The correct unit of analysis is paying usage relative to total cost, including power, depreciation, networking, and model training.
Where listed stocks fit
A new model-company contract might help a chip vendor immediately while creating a long payback period for a cloud operator. Power and cooling suppliers depend on infrastructure delivery, while cloud returns depend on sustained paid usage. Data-center orders can lead realized end-customer demand by years.
What could change the view
Rising paid usage, enterprise retention, higher utilization, and falling cost per useful task would strengthen the investment case. Fixed equipment and power commitments ahead of profitable demand would weaken it. The decisive measures are AI revenue and operating cash relative to capex and depreciation, timely grid connections, and supplier backlog conversion at sustained margins.
Current view: The demand signal and funding are large; attractive returns across the entire AI buildout remain unproven. Published company results and physical power delivery matter more than announced capacity totals.
Sources
- U.S. Department of Energy: 2024 U.S. data-center energy-use study — historical load and 2028 scenario range.
- OpenAI: March 2026 financing and operating claims — issuer statements on funding, revenue pace, and partners.
- Anthropic: May 2026 Series H announcement — issuer funding and demand claims.