Hyperscaler Capex: $300B at Alphabet, $175B at Microsoft — What Demand for Datacenters?

The quarterly results just released give the AI infrastructure cycle rare visibility. Microsoft is targeting roughly $175 billion of capital expenditure for calendar year 2026, after a quarter in which its capex nearly doubled to $39.4 billion. Alphabet is preparing more than $300 billion for 2027. Behind these figures lies a physical reality: datacenters to build, connect and operate.

Demand is clearly there: Azure is growing 48% year on year, Microsoft Cloud reached $63.9 billion in quarterly revenue, and Alphabet's cloud doubled to $30.7 billion. These are no longer bets on future AI usage — they are existing cloud revenues pulling capacity.

The stock market, however, is discerning: Alphabet fell as much as 10% after its results, penalized for spending deemed high despite growth. Investors no longer question demand; they scrutinize returns. This is an important nuance for anyone investing in the infrastructure chain.

Multi-Year Demand, Not a Flash in the Pan

What these guidances change is the visibility horizon. Capex committed for 2026 and 2027 means capacity orders materializing over several years: servers, networks, but also buildings, grid connections and cooling. The physical infrastructure chain benefits from an order-book effect.

For Europe, the question is how much of these gigantic envelopes will land on the continent. Digital sovereignty, data residency and latency requirements argue for a growing share of European capacity — provided sites, permits and megawatts are available when orders arrive.

What This Changes for Investors

The first reading is favorable: customers whose capex runs into hundreds of billions are solid counterparties and signatories of long contracts. Datacenter demand no longer rests on projections, but on voted and published budgets.

The second reading calls for discipline: the market reaction to Alphabet reminds us that investors' patience has limits. If AI returns disappoint, capex will be revised — and capacity most exposed to a single customer or short contracts will suffer. Selectivity on counterparties and commitment duration remains the best protection.

How We Apply This at FINXIA

In the TITAN DC AI strategy, these guidances support our central scenario: the scarcity of prepared sites and connected megawatts will persist against multi-year demand. We favor assets whose value lies in what hyperscalers cannot accelerate — connected land, available power, permits.

We also retain the lesson of the market reaction: our counterparty analysis distinguishes firm commitments from intentions. An asset backed by a long contract with a customer whose capex is published and financed does not have the same profile as speculative capacity built on expectations.

Key Takeaways

Microsoft is targeting roughly $175 billion of capex in 2026 and Alphabet more than $300 billion in 2027, with cloud growth of 48% and a doubling respectively. AI infrastructure demand is documented and multi-year — but markets now demand proof of returns, which strengthens the premium for well-contracted assets.

Worth watching: the European localization of this capex, upcoming capacity contracts signed in Europe, and how hyperscalers discuss AI return on investment in coming quarters.