Our position
The AI build-out is now one of the largest capital programs in corporate history, and it is running into physical limits: power, cooling, grid connections and the balance sheets that pay for it. We believe the most durable startup opportunities in 2027 sit where those limits bind. The question is no longer only who has the most GPUs, but who can turn a megawatt into the most useful computation.
The scale of the build-out
According to FactSet, Alphabet raised its 2026 capital expenditure guidance to $180 billion to $190 billion, and Meta raised its guidance to $125 billion to $145 billion. Across the largest cloud and AI spenders, calendar 2026 capex guidance is close to $800 billion including finance leases and customer prepayments, with cash capex for fiscal 2026 expected to exceed $690 billion. That is growth of more than 80% year over year, after roughly 70% the year before.
The financing is changing too. FactSet reports that incremental debt as a share of capex rose from 9% in fiscal 2024 to 32% in the twelve months to June 2026, taking aggregate debt for the group to about $700 billion, and that free cash flow is expected to approach zero or turn negative this fiscal year for all of them except Alphabet and Microsoft.
| Indicator | Reading | Source |
|---|---|---|
| Alphabet 2026 capex guidance | $180B–$190B | FactSet, Jul 2026 |
| Meta 2026 capex guidance | $125B–$145B | FactSet, Jul 2026 |
| Hyperscaler calendar 2026 capex guidance | Close to $800B incl. leases | FactSet, Jul 2026 |
| Debt as a share of incremental capex | 32% (vs 9% in FY24) | FactSet, Jul 2026 |
| Global data centre electricity use | ~415 TWh (2024) to ~945 TWh (2030, projected) | IEA, Energy and AI |
Why power is the constraint
The International Energy Agency projects that electricity consumption by data centres will roughly double by 2030, to around 945 terawatt-hours, driven mainly by AI. New capacity needs generation, transmission and interconnection, all of which move on utility and permitting timelines measured in years. Chips can be bought in quarters. Power cannot.
Venture investors are responding. Crunchbase reports that in Q3 2026 data centers, semiconductors and energy each attracted more than $10 billion of venture funding, and that AI infrastructure companies such as Crusoe and Nscale raised rounds of more than $3 billion.
Where we see opportunity
- Power at the site: on-site and behind-the-meter generation, storage and grid-interactive data centres that can flex load to win faster interconnection.
- Efficiency per watt: inference optimization, scheduling, model compression and workload placement that raise useful output per megawatt.
- Cooling and thermal systems: liquid and immersion cooling, heat reuse and the components that make high-density racks practical.
- Grid software: interconnection, forecasting and planning tools that shorten the time from site selection to energized capacity.
- Infrastructure finance and operations: tools that help developers, lenders and operators underwrite and run capital-intensive assets.
What we are cautious about
- Businesses whose returns depend on today's GPU rental prices persisting.
- Single-customer models exposed to one hyperscaler's capex decisions.
- Hardware plans that need project-scale capital before the technology is proven. Our research on financing deep tech sets out how to sequence that capital.
What we want to see from founders
- The physical constraint you remove, and how you measure it: megawatts, months, dollars per token.
- Who pays you, and why they will keep paying if capex slows.
- A capital plan that matches each stage of risk to the right kind of money.
Frequently asked questions
How much are hyperscalers spending on AI in 2026?
FactSet reports that calendar 2026 capex guidance across the largest cloud and AI spenders is close to $800 billion including finance leases, with Alphabet guiding $180–190 billion and Meta $125–145 billion.
Why is power the bottleneck for AI?
New data centre capacity needs generation, transmission and grid interconnection, which take years to build, while chips can be bought within quarters. The IEA expects data centre electricity use to roughly double by 2030.
What AI infrastructure startups does Cash Capital look for?
Companies that remove a physical constraint, such as power, cooling or interconnection time, and can measure it in megawatts, months or cost per unit of computation.
Sources
- FactSet: Hyperscalers tap external financing as AI capex outruns cash flow
- IEA Energy and AI report coverage, via The Star
- Crunchbase News: Q3 2026 global startup funding, billion-dollar rounds and exits
Important information. This publication is provided for general information only and reflects the views of Cash Capital as of its date. It is not investment, legal, tax or accounting advice, and it is not an offer to sell, or a solicitation of an offer to buy, any security or any interest in any fund.
Figures, statistics and company information are drawn from third-party sources that we believe to be reliable, including news reports, data providers and company announcements. We have not independently verified them, and some may be estimates, preliminary, reported rather than confirmed, or later revised. Cash Capital makes no representation or warranty as to their accuracy, completeness or timeliness and accepts no liability for any loss arising from reliance on them. Readers should consult the original sources and their own advisers before making any decision.
Predictions, outlooks and other forward-looking statements are opinions about uncertain future events and may not occur. We do not undertake to update this publication. Companies named are used for illustration only; unless stated otherwise they are not Cash Capital portfolio companies, and naming them does not imply any relationship, endorsement or recommendation. See our Terms & Disclosures.