13JUN

Big Five to spend $650bn on AI in 2026

3 min read

11:22UTC

The five largest US technology companies plan to nearly double AI infrastructure spending in 2026, converting payroll budgets into data-centre capacity at a pace that locks in years of automation pressure.

← Back to Washington pulls a live AI model Jump to analysis ↓

EconomicAssessed

Key takeaway

Private AI capex now rivals the entire US interstate highway system's historical cost — in a single year.

The five largest US technology companies plan to spend $650–690 billion on AI infrastructure in 2026, nearly doubling their combined outlay from the previous year, according to a Bridgewater Associates estimate ¹.

The capital flows into data centres, GPU procurement, and power infrastructure. Meta's capex guidance and Oracle's planned workforce-to-infrastructure conversion are two expressions of a sector-wide pattern: labour budgets becoming infrastructure budgets at accelerating rates ².

The scale creates its own momentum. Data centres take two to four years to plan, permit, and build. They consume electricity at densities far exceeding traditional computing, adding grid constraints to the capital lock-in. Once tens of billions are sunk into physical infrastructure, the economic incentive to automate enough work to justify the investment intensifies. The capital demands utilisation, which means finding more tasks to transfer from workers to machines. The current wave of layoffs is the front end of a capital cycle that will generate sustained pressure on labour costs through the rest of the decade.

Deep Analysis

In plain English

Five companies are collectively planning to spend more on computer infrastructure in 2026 than the entire US government spends on defence procurement. This money flows primarily to specialised chips (GPUs), the vast warehouses housing them (data centres), and the electricity to run them. It does not flow to hiring more workers — these same companies are simultaneously cutting headcount. The bet is that AI will make each remaining worker so much more productive that the economics work out. Whether that bet is correct determines whether this is the largest productive investment in private-sector history or the largest capacity overbuild.

Deep Analysis

Synthesis

This $650–690B figure represents a privately funded reorientation of the US capital stock at a speed with no peacetime precedent. Gains accrue to a narrow set of capital owners — GPU manufacturers (primarily Nvidia at 75%+ gross margins), construction firms, and energy utilities — while the labour market contracts. This is capital deepening at wartime mobilisation speed, but without the corresponding employment surge that wartime investment historically produced.

Root Causes

The body frames this as AI-driven, but a structural factor it omits is that hyperscaler cloud revenue is itself growing rapidly, creating internal compute demand that is partly independent of AI product revenue. AWS, Azure, and Google Cloud are building infrastructure for paying cloud customers; AI is an accelerant layered onto an existing secular trend rather than the sole cause.

Escalation

The spending commitment is largely locked in for 2026 through multi-year data-centre construction contracts and GPU supply agreements. Even if AI revenue disappoints, the capex will be spent — creating a sunk-cost dynamic that may extend the investment cycle beyond rational return thresholds.

What could happen next?

1 risk2 consequence1 precedent1 meaning

Risk
If AI revenue fails to materialise at projected scale, sunk construction and GPU contracts create a capacity overbuild with no viable exit mechanism.
Medium term · Suggested
Consequence
Nvidia captures the dominant share of capex at 75%+ margins, concentrating wealth gains more narrowly than any comparable historical infrastructure boom.
Short term · Assessed
Precedent
Hyperscaler monopoly over AI infrastructure may invite utility-style regulation analogous to interventions that followed railway and telecom concentration.
Long term · Suggested
Consequence
Residential electricity bills in data-centre-heavy regions face upward pressure as utility load growth is passed through to consumers.
Short term · Suggested
Meaning
Capital deepening at this pace without employment growth inverts the historical relationship between investment booms and job creation.
Medium term · Assessed

Source Landscape

This story draws on mixed-leaning sources from United States and United Kingdom

United States

United Kingdom

LeftRight

Primary parallel: The US railway construction boom of 1847–1857 deployed private capital at unprecedented GDP-relative scale into infrastructure, producing monopoly concentration, the financial panic of 1857, and eventual regulatory intervention via the Interstate Commerce Commission in 1887. Critically, the companies that built the rails rarely captured the full economic surplus — downstream industries benefited more.

Counter-parallel: The 1990s telecom infrastructure buildout (over $1 trillion in fibre and network investment) produced genuine long-run productivity gains but also massive overcapacity and severe write-downs before the assets became economically productive. The benefit materialised eventually, but investors in the infrastructure itself suffered first.

Why it’s relevant now

If the railway pattern holds, AI infrastructure investors may not be primary beneficiaries of the productivity gains they are funding. The regulatory intervention that followed railway monopoly also foreshadows potential antitrust or utility-style oversight of hyperscaler AI infrastructure.

Nvidia's gross margin on data-centre GPUs exceeds 75%, meaning a substantial fraction of this $650–690B flows directly to a single supplier's profit rather than circulating through the broader economy. This is structurally different from historical infrastructure booms where spending dispersed across materials, labour, and regional contractors.

The electricity sector stands to capture an estimated $40–60B in additional annual revenue from AI data-centre load by 2027. Regulated utilities in data-centre-heavy states face an unusual dynamic: load growth that benefits shareholders but requires rate increases passed to residential consumers, creating a regressive distributional effect.

Consensus view: Goldman Sachs Global Investment Research estimates AI data-centre build-out will drive US electricity demand growth of 160 terawatt-hours annually by 2030 — roughly equivalent to adding a state the size of California to the grid.

Counter-view: Researchers at China's State Council Development Research Centre argue that concentrated Western AI infrastructure investment creates strategic dependency risk for non-Western nations unable to match this capex scale, potentially entrenching a new form of technological dependency analogous to the 19th-century industrial gap.

Key tension: Whether $650–690B in capex will generate commensurate productivity and revenue gains, or constitutes the largest capacity overbuild in technology history, is the central unanswered empirical question.

Sources:Bloomberg·Reuters

Mentions:China →Iran-Azerbaijan Conflict (2026) →Goldman Sachs →OpenAI →Iran-Israel-US Conflict 2026 →Meta →

First Reported In

Update #1 · Meta cuts 20% while Big Tech spends $650bn

Bloomberg· 17 Mar 2026

Read original →

Causes and effects

This Event

Big Five to spend $650bn on AI in 2026

At $650–690 billion, the committed capital creates a self-reinforcing cycle — data centres that take years to build and demand utilisation, generating ongoing economic incentive to automate work to justify the investment.

Led to

AI capex binge drains Big Tech cash flow

The $650-690B AI infrastructure spending wave is driving the FCF declines at Meta and Microsoft forecast by Barclays

Occurred 5 Feb 2026

Read story →

Azure up 40%, Microsoft posts $77.7bn

Microsoft's Q1 FY2026 results were the first financial data point testing whether Big Five AI infrastructure commitments of $650-690 billion would generate commensurate revenue.

Occurred 26 Mar 2026

Read story →

Oracle cuts 30,000 to fund AI

The $650-690 billion Big Five AI capex commitment created the competitive pressure driving Oracle 156 billion infrastructure bet.

Occurred 31 Mar 2026

Read story →

Different Perspectives

India IT services and global capability centre workforce

India's in-house GCCs added roughly 200,000 net staff in fiscal 2026, nearly double the 110,000 added by the IT services firms feeding the same companies. The shift moves work toward captive centres while squeezing entry-level hiring at the outsourcing firms, reshaping where Indian tech careers begin as US clients cut staff at home.

EU workers and European labour institutions

The 93-4 committee vote locked the diluted Omnibus literacy clause before plenary: EU workers in AI-augmented but non-high-risk workplaces have no statutory right to demand an explanation until December 2027. The European Trade Union Confederation called the shift from 'ensure' to 'support' a legal threshold collapse, not a drafting compromise.

UK workforce and labour market

UK 16-to-24 unemployment reached 16.2% in the latest ONS reading, above the 15.2% pandemic peak and the highest since 2015. Britain is among the most AI-exposed labour markets this desk tracks, yet the Office for National Statistics still publishes no AI-attribution layer, so young workers face the displacement without official data naming its cause.

Anthropic and frontier AI labs subject to US jurisdiction

Anthropic complied with the directive but publicly disputed its application, citing that OpenAI's GPT-5.5 carried the identical jailbreak vulnerability and remained on sale. For any US-domiciled frontier lab, the action demonstrates that regulatory compliance and political alignment are now distinct variables: Anthropic backed the pro-regulation PAC and was the first lab Washington reached.

US national-security and export-control apparatus

The Lutnick directive treats runtime inference access by a foreign national as legally equivalent to exporting Claude Fable 5 and Mythos 5 to that person's home country. It established that a deployed consumer AI product can be withdrawn globally by regulatory letter, with no appeal period and no customer notice.

European workers and regulators

NBER working paper w34995 found European workers use generative AI at 32% versus 43% of US workers, a gap driven by management practice rather than regulation. The EU AI Act's high-risk employment deadline stays at December 2027, leaving European workers facing the same displacement curve two to four years behind the US.