Tag: data centers

Germany and France put digital sovereignty into operational terms
Germany and France have published a joint paper on digital sovereignty, dated 17 June 2026. It is only six pages long, but it does something useful: it gives the term digital sovereignty a set of testable criteria.

Europe has spent years talking about sovereignty in broad terms. The Franco-German paper asks a narrower question: when a government, company or public institution buys digital technology, what would make that technology more or less sovereign?

The paper does not pretend this is easy. It says digital sovereignty should be risk-based, modular and scalable. It avoids protectionism and isolation. It leaves defence and national security outside its scope. It creates no direct budget obligation and does not impose conditions on private procurement.

The document is cautious by design. That is useful for consensus. It is also the problem.

Germany and France are not proposing a simple "buy European at any cost" doctrine. They are proposing criteria that could feed into the EU Tech Sovereignty Package, including the Cloud and AI Development Act. If those criteria survive the legislative process, they could start shaping procurement, cloud architecture, sensitive-data handling and public-sector technology choices.

The paper's value is the checklist. Its weakness is that it stops there. It does not yet create the kind of aggressive investment push now visible in other regions.

The definition is broader than cloud

Digital sovereignty has to be assessed across the stack, from chips and networks to cloud platforms and AI.

The core definition is worth reading carefully. Digital sovereignty is described as the capability and capacity to develop, provide, use, adapt and control digital technologies, including hardware, in an independent, self-determined and secure manner.

Data location is only one part of it.

It includes hardware, software, data handling, AI, semiconductors, cloud, quantum, robotics, cybersecurity, standards, supply chains, skills and control over operational processes. The paper says critical dependencies exist across the entire stack, from IT infrastructure and semiconductors to software, data and AI.

This maps better to how dependency actually works.

Europe's dependency problem is scattered across the stack: hyperscale cloud, chips, operating systems, cybersecurity tools, AI models, productivity platforms, data infrastructure, technical standards, venture capital depth, and the ability to scale startups into global companies.

One datapoint stands out: in Europe's digital industrial ecosystem, most companies have fewer than 250 employees, based on the European Commission/JRC SME report cited in the paper. That captures one of Europe's structural problems. Europe has plenty of innovation. It has too few digital companies with global scale.

The six criteria matter most

The six criteria can be used in procurement, supplier reviews, architecture decisions and exit planning.

The paper defines six dimensions of digital sovereignty.

The first is the capability to implement and enforce. This is about whether Europe can apply its own legal and security conditions in practice. The criteria include EU-law compliance, transparency of ownership and subcontractor chains, disclosure of dependencies on third countries, restriction of sovereignty-critical extraterritorial data access, and the ability to investigate cybercrime and state-backed attacks.

The cloud debate often gets stuck here: legal jurisdiction and operational control do not always sit in the same place as the data center.

The second is the capability to design, deploy and use technologies. This includes scientific ecosystems for AI, microelectronics, robotics, data, quantum and cybersecurity; industrial demand for key technologies; research transfer; startup scaling; open source, open hardware and interoperability; and participation in standardisation.

Europe often underestimates this layer. Regulation can define the rules. It cannot replace the people, companies and institutions that build, operate, buy and improve the technology.

The third is economic value creation. The paper looks at where value is generated: R&D, engineering, skilled employment, operational control and contribution to the European technology ecosystem. It also explicitly allows partial value creation in trusted partner countries. That keeps the framework open enough to be economically realistic.

The fourth is protection of data. The paper calls on the European Commission to define the highest protection standards for the most sensitive data, including safeguards against cybersecurity risks and the effects of non-EU extraterritorial legislation. It also mentions mandatory privacy-enhancing technologies.

Sensitive data policy is now also industrial policy.

The fifth is substitutability and interoperability. The paper asks for modular architecture, open standards, open interfaces, software bills of materials, migration paths, exit concepts and multi-vendor strategies. In plain English: do not build systems that cannot be changed later.

For me, this is the most practical part of the paper. Lock-in rarely arrives as a crisis. It arrives as a procurement decision that cannot be reversed without years of cost and disruption.

The sixth is infrastructure resilience. The paper calls for sovereign data centers, AI, quantum and cloud computing infrastructure, interchangeable hardware and software stacks, diversified supply chains, secure and sustainable energy, high-performance networks and access to critical space resources.

Digital sovereignty depends on the physical layer too: data centers, energy supply, networks and resilience.

This links directly to the SoftBank France data-center story. Digital sovereignty now has a power, land, data-center and network dimension. The debate has moved well beyond data location and cloud labels.

The paper is careful, maybe too careful

The paper is politically careful. It is non-binding. It excludes defence and national security. It does not force public spending. It does not impose rules on private procurement. It stresses trade obligations, trusted partners and cost efficiency.

That makes it weaker than a real industrial plan. It also makes the document harder to dismiss as protectionism.

The gap is not definition. The gap is action.

The paper does not unlock capital. It does not create major public procurement demand. It does not accelerate data-center buildout, AI infrastructure, semiconductor capacity, cloud scale or startup growth. It gives Europe a framework for assessing sovereignty, but it does not yet give European providers the demand, reference customers or balance-sheet confidence needed to scale.

The paper does not argue for closing Europe off. Its more useful move is to make dependency measurable. Who owns the provider? Which subcontractors matter? Where is R&D located? Can the customer exit? Are open interfaces available? Can sensitive data be protected from extraterritorial access? Can Europe still operate if one supplier, jurisdiction or supply chain becomes unavailable?

These questions belong in procurement files, architecture reviews and risk discussions.

For enterprise leaders, digital sovereignty is becoming a procurement and architecture discipline. It will affect cloud strategy, AI deployment, data classification, supplier concentration, cybersecurity, exit planning and board-level risk.

For policymakers, a definition is useful only if it changes incentives. Europe needs procurement demand for sovereign solutions, faster scaling paths for startups, deeper capital markets, serious public-sector reference customers, and infrastructure policy that connects cloud, AI, energy, semiconductors and networks.

Without that, sovereignty stays a vocabulary exercise. Other regions are moving with capital, infrastructure, industrial policy and large anchor customers. Europe cannot answer that with criteria alone.

The executive takeaway

The Franco-German paper stops short of a sovereignty plan. It offers criteria. Criteria still matter because they shape what governments and large buyers start asking for. They shape tenders. They influence compliance teams. They tell suppliers what the next market standard may look like.

If Europe uses this framework well, sovereignty becomes less abstract: fewer lock-ins, clearer exit paths, more transparent supply chains, stronger data protection, more European value creation, and better infrastructure resilience.

If Europe uses it badly, it becomes another vocabulary layer on top of slow procurement and fragmented national initiatives.

My read: this paper is strongest where it is most practical. It connects sovereignty to ownership, enforceability, interoperability, data protection, value creation and infrastructure. It avoids the fantasy of full autarky. It accepts trusted partners. It treats sovereignty as a risk-based capability, not as a flag on a server.

But the next test is not another definition. It is demand.

Without procurement demand, budgets, infrastructure, reference customers and scale, European providers will stay small. Without scale, the dependency problem stays exactly where it is.

Bottom line: good start. Now Europe needs action.

Sources and further reading
- Franco-German Joint Paper on Digital Sovereignty, 17 June 2026
- BMDS: Germany and France present joint definition of digital sovereignty
2026-06-21
AI Funding Is Turning Into Infrastructure Capital
Crunchbase‘s April report reads, at first, like one more data point in the AI boom. Global venture funding hit $56 billion in April 2026 – the third-biggest month in a year, and roughly double April 2025. AI took $37 billion of that, about two-thirds of all venture money in the month.

What matters is where the money went. Two rounds did most of the work. Anthropic raised $15 billion. Jeff Bezos’s Project Prometheus, aimed at AI for manufacturing and the physical world, raised $10 billion. Together they accounted for 45% of all venture funding in April. Five weeks later, on 28 May, Anthropic closed a $65 billion Series H at a $965 billion valuation – the largest equity round ever raised by an AI company, and enough to pass OpenAI as the most valuable startup in the world.

These rounds work differently from the software rounds that came before them. Venture capital has started to behave like strategic industrial capital, and the AI race has become a contest over who can assemble enough capital, compute, power, data, and industrial access to own the next operating layer of the economy.

The money is pooling at the top

The headline funding number can rise while the market underneath it narrows.

Venture has always followed a power law: a few companies take most of the returns. April pushed that to an extreme. Through April, global venture investment was up 139% year over year, and nearly 60% of that capital went to just five companies – most of them backed by cash-rich public tech firms, private equity, and the largest VC funds. Q1 looked the same: OpenAI ($122 billion at an $852 billion valuation), Anthropic, xAI, and Waymo took roughly two-thirds of all global venture funding between them.

This changes what the funding totals tell you. In an ordinary cycle, rising funding signals broad risk appetite – more founders backed, more categories opening, more experiments running. Right now the total can climb while the market narrows underneath it. Plenty of money is flowing, but it reaches very few companies, and the ones it reaches have started to look like national-scale infrastructure projects.

That is why the comparison to past SaaS or internet cycles falls apart. A $15 billion AI round belongs to an entirely different category of capital formation than even the largest software growth round.

Models have become capital assets

A frontier model is no longer just an algorithm. It is a capital asset tied to compute, chips, cloud and distribution.

AI model companies raised $26.7 billion in April – by far the largest single category, ahead of physical AI ($5.3 billion) and AI infrastructure like chips and data centers ($1.8 billion).

The reason is structural. Frontier labs are expensive in ways software companies never were: they need long compute contracts, data-center capacity, advanced chips, large engineering and safety teams, enterprise sales, and deep ties to the hyperscalers. They sell software and spend like heavy industry.

The cloud era made infrastructure feel weightless. You rented compute, scaled on demand, and built globally without owning anything. AI has partly reversed that. Compute has turned back into a scarce, physical input that decides who can compete, so the companies with privileged access to chips, power, and distribution hold a real structural edge. That is why hyperscalers, sovereign funds, and private equity keep moving closer to the center of AI financing.

Anthropic‘s Series H is the clearest example. Look at who funded it: alongside the crossover investors sit the companies that supply the infrastructure Claude runs on – the cloud it trains on, the memory chips that serve its inference. Those backers have a direct operating interest, since their own businesses grow as Anthropic grows. A model company has become a capital asset that its own suppliers want a stake in.

Physical AI is the second signal – and maybe the bigger one

Physical AI shifts the question from digital productivity to industrial leverage.

The Prometheus round may matter more than Anthropic‘s, even though it is smaller. Anthropic represents the frontier-model race. Prometheus points to the phase after it: AI moving out of language and code and into engineering, manufacturing, robotics, aerospace, automotive, and physical production. Crunchbase counted about $5.3 billion of April’s AI funding as physical AI – a small slice today, with an outsized claim on the real economy.

For a few years, AI has mostly been a knowledge-work story: it writes, summarizes, codes, plans, and automates digital tasks. The physical-AI bet says the next contest is over the industrial system itself – compressing engineering cycles, simulating physical systems, optimizing factories, improving robotics, speeding up materials discovery. If that works, the real value sits in industrial leverage: how quickly companies can design, test, and build physical things.

That also explains the capital intensity. Industrial AI demands labs, data rights, robotics environments, manufacturing partners, domain experts, and access to the messy operational data inside real companies. The winner here will probably be whoever can wire models into real factories, supply chains, machines, and the proprietary data that sits inside them.

Public and private markets are now one loop

The April data also shows how tightly public markets, private markets, and the wider economy are now linked. Alphabet, Microsoft, and Amazon all beat revenue expectations while spending heavily on AI infrastructure. Pantheon Macroeconomics estimates that about half of the 2% U.S. GDP growth in Q1 came from AI buildout. That figure is large enough to matter: AI now shows up directly in the macro data.

The result is a feedback loop. Public tech companies throw off cash and market value. Those balance sheets fund compute and strategic investments. The investments flow into private AI companies, which buy more infrastructure, which lifts hyperscaler revenue and capex again. For now, the loop is strong.

The risk is that it makes AI look broader than it is. When a few capital-rich companies drive both the public-market narrative and the private-market totals, the whole ecosystem leans on a small set of balance sheets and assumptions. The boom is genuine, and it is also concentrated, circular, and dependent on a narrow base of infrastructure.

What this means for Europe

U.S. companies raised $39 billion in April, around 70% of global venture funding. For Europe, the clean comparison is not AI-only funding; it is total venture/startup funding on the same monthly basis. A Crunchbase-based European VC landscape dataset counted $4.8 billion across 327 European investments in April, while Tech.eu counted €5.1 billion across 290 European tech deals. Even allowing for methodology differences, Europe was roughly a one-tenth-of-global market while the U.S. took about 70%. That should sting.

The usual European AI debate is about regulation, foundation models, talent, data, and digital sovereignty. All of it matters. April adds a dimension that gets less attention: capital sovereignty. If AI leadership now takes tens of billions for models, data centers, chips, power, and industrial deployment, then good research and sensible rules will not be enough on their own. Europe also has to mobilize capital at the scale and speed the technology demands.

This is where the Draghi competitiveness argument gets concrete. Europe cannot regulate its way to AI relevance, and it cannot research its way there either while its capital, compute, and adoption stacks stay fragmented.

The position is far from hopeless. Europe has real industrial depth – manufacturing, automotive, aerospace, energy systems – in exactly the domains where physical AI could matter most. That strength does not convert into AI advantage automatically. It has to be connected to capital, compute, data-sharing arrangements, procurement, and faster decisions. Otherwise the industrial data and engineering know-how that should be Europe’s edge will be monetized through platforms funded and controlled elsewhere.

The question for leaders

For executives, the useful question is what kind of market is being built, and whether their company has a place in it. If AI funding is becoming infrastructure capital, then AI strategy belongs in the boardroom as a question about strategic dependency:
- Who controls the models you rely on?
- Who controls the compute?
- Who owns the industrial data?
- Who has the capital to build at scale?
- Who can turn AI capability into operating-model change faster than you can?
This matters most for companies outside tech. Many industrial, financial, logistics, healthcare, and public-sector organizations still treat AI as a vendor-selection exercise, and that framing is too small. The real question is where you sit in the emerging AI capital stack – as a buyer of capability, a supplier of domain data, a deployment partner, a regulated adoption environment, a business whose workflows get compressed by someone else’s model, or a company that uses AI to redesign the economics of its own industry.

What I’m watching next

Three signals matter more than the next monthly funding total.
1. Concentration. If capital keeps pooling in a few frontier-model and infrastructure companies, the AI market will increasingly resemble a strategic infrastructure race.
2. Physical AI. If funding for robotics, manufacturing, and autonomy accelerates, AI starts reshaping the industrial economy, well beyond office work.
3. Europe. If the continent stays strong on regulation and weak on capital mobilization, the sovereignty debate stays rhetorical.
April’s data points to an AI economy that is becoming more capital-intensive, more concentrated, and more physical. The next phase will be won by whoever can put the full stack together: capital, compute, energy, data, industrial access, distribution, and execution speed. That is a different kind of technology race, and it is already running.

Sources: Crunchbase, “Billion-Dollar AI Rounds Push April To Third-Highest Startup Funding Month In A Year” (5 May 2026) and the Q1 2026 global funding report; Trustventure, “European Venture Capital Landscape – April 2026”; Tech.eu, “April 2026’s top 10 European tech deals”; Anthropic’s Series H announcement and reporting from Axios, CNBC, TechCrunch and Fortune (28 May 2026); GDP estimate from Pantheon Macroeconomics.

Sources and further reading
2026-05-31
AI’s next bottleneck may not be intelligence. It may be Earth.
For the last two years, the AI debate has been mostly about intelligence.

Which model is ahead? How fast are capabilities improving? Will agents replace tasks, jobs, or whole workflows? Can Europe regulate the technology fast enough?

All valid questions.

But the next constraint may be less abstract. It may be physical.

Power. Grid capacity. Land. Cooling. Permits. Transmission lines. Water. Construction time. Capital allocation.

The AI race is turning into a gigawatt race. And if the space-data-center discussion is any signal, the next frontier may not just be cloud regions. It may be orbit.

My read: the executive conversation has to move from "Which AI model should we use?" to "What physical infrastructure does our AI strategy depend on?"

The scale shift

The scale jump matters: 10 MW is a facility, 100 MW is industrial infrastructure, and 1 GW becomes a regional energy strategy.

A modern hyperscale data center is not a large office building with servers. It is an industrial energy asset.

The International Energy Agency says average data centers draw around 5-10 megawatts. Large hyperscale facilities increasingly require 100 megawatts or more. That number sounds technical, so translate it.

One megawatt running continuously for a year equals 8.76 gigawatt-hours. A 100 MW data center therefore consumes 876 GWh per year, or 0.876 TWh. At 90% utilization, still roughly 0.8 TWh per year. The IEA compares this to the annual electricity demand of about 350,000 to 400,000 electric cars.

A 1 GW AI campus is ten 100 MW hyperscale data centers. Running continuously, it consumes 8.76 TWh per year.

For comparison, Germany's annual electricity consumption is roughly 500 TWh. The EU is around 2,700 TWh. The US is around 4,000 TWh. So one 1 GW AI campus would be small at continental scale – about 0.3% of EU electricity consumption or 0.2% of US consumption – but huge at local grid scale.

That local point matters.

Put a 1 GW load in the wrong county, with weak transmission and slow permitting, and it is not "0.2% of America." It is a grid emergency, a political fight, and a capital allocation problem.

Now consider the language around terawatts. Elon Musk's recent "Terafab" discussion was about chip manufacturing, not a conventional data center, but the vocabulary matters. AI infrastructure ambition is moving from mega to giga to tera. A theoretical 1 TW compute or manufacturing footprint running continuously would consume 8,760 TWh per year. That is more electricity than the US and EU combined.

That does not mean a 1 TW data center is around the corner. It means the ambition curve is now colliding with the energy system.

The current footprint

The IEA estimates global data center electricity consumption at 240-340 TWh in 2022, excluding crypto mining. That was around 1-1.3% of global final electricity demand.

In large economies such as the United States, China and the European Union, data centers already account for around 2-4% of total electricity consumption. That is the average.

The local reality is more extreme.

The IEA notes that data centers have already surpassed 10% of electricity consumption in at least five US states. In Ireland, data centers account for more than 20% of electricity consumption. Denmark projects data center electricity use could rise sixfold by 2030 and approach 15% of national electricity consumption.

This is the important distinction: globally, data centers are still a manageable share of electricity. Locally, they can become one of the dominant loads on the system.

Goldman Sachs Research estimates data center power demand could grow 160% by 2030, with global data centers rising from roughly 1-2% of power consumption today to 3-4% by the end of the decade. It also estimates AI could add around 200 TWh per year of data center power demand between 2023 and 2030.

Two hundred TWh is not abstract. It is close to the annual electricity consumption of a mid-sized industrial country. And it is only the AI-related increment in one forecast.

The backlash is already here

Global averages hide local pressure: data centers can reach double-digit shares of electricity demand in specific regions.

This is no longer theoretical.

In May, several local flashpoints showed the political side of the bottleneck. Seattle was weighing a pause on large data centers. Durham, North Carolina passed a 60-day moratorium on data-center development. A Texas county paused data-center construction in rural areas for a year. Utah approved a data-center project described as twice the size of Manhattan, triggering backlash. Tennessee was considering legislation that would let data centers self-power with limited regulation.

Different places, same pattern.

AI infrastructure is colliding with local politics. Communities are asking who gets the jobs, who pays for grid upgrades, who carries water risk, who absorbs noise and land-use impact, and who benefits from the compute.

This is the part of the AI story many executives still underestimate. It is not enough to have GPUs. You need permission. You need interconnection. You need credible energy sourcing. You need community acceptance.

The future of AI may be decided as much in planning boards and utility queues as in model labs.

Why energy is now part of AI leadership

AI energy strategy is now an executive checklist: economics, thresholds, model allocation, partnerships, and efficiency.

For a long time, digital leaders could assume infrastructure would scale behind the scenes. Cloud abstracted away servers. SaaS abstracted away operations. Developers increasingly acted as if compute was infinite, elastic, and mostly someone else's problem.

AI breaks that illusion.

Training frontier models is energy-intensive. Inference at scale may matter even more because successful AI products are used continuously. Agents add another multiplier: they do not just answer one prompt. They plan, call tools, retry, search, generate, check, and act. A single user request can become dozens or hundreds of model calls behind the scenes.

That makes energy not just an engineering issue but a leadership issue.

If AI becomes a core production layer, power becomes part of product economics. Latency becomes part of geography. Energy procurement becomes part of risk management. Infrastructure partnerships become part of market entry. Sustainability claims become harder to defend if absolute consumption rises faster than efficiency improves.

The better question is not whether AI uses "too much" energy.

The better question is: are we using scarce energy for high-value intelligence, or are we wasting it on low-value automation theatre?

The opportunity

The upside is enormous.

AI can help design better grids, forecast demand, optimize industrial processes, improve cooling, accelerate materials science, reduce waste, and make energy systems more flexible. The same technology that increases electricity demand can also improve how electricity is produced, routed, stored, and consumed.

There is also a market opportunity.

Companies that solve the infrastructure layer will not just be suppliers to AI. They will become strategic gatekeepers. Power developers, grid operators, data-center builders, cooling specialists, chip designers, construction firms, nuclear developers, storage providers, and energy software companies are moving closer to the center of the AI economy.

This is especially relevant for Europe.

Europe often frames AI competitiveness around regulation, foundation models, sovereignty, and talent. All matter. But infrastructure sovereignty may become just as important. If compute depends on power availability, grid speed, and data-center capacity, then AI sovereignty is partly electricity sovereignty.

A European AI strategy without an energy strategy is incomplete.

The space question

Space-based data centers are not a near-term replacement for terrestrial infrastructure. They are a signal that the AI compute curve is pushing beyond the grid.

The more provocative version of this debate is space.

A few years ago, data centers in orbit sounded like science fiction. Now Bloomberg is writing about how to build them. McKinsey has made the case for space-based data centers. University researchers are exploring the idea because AI energy demand is rising. Google and SpaceX have been linked in recent coverage to the broader possibility of AI data centers in space.

The attraction is obvious: continuous solar power, less terrestrial land pressure, potentially easier cooling through radiative systems, and the strategic appeal of moving part of the compute layer off Earth.

The problems are just as obvious: launch cost, maintenance, radiation, latency, orbital debris, security, regulation, and basic economics.

But the fact that serious people are asking the question matters. Space data centers are not a near-term replacement for terrestrial infrastructure. They are a signal. The AI compute curve is steep enough that people are looking beyond the grid.

When a technology forces executives to ask whether the data center belongs in orbit, something fundamental has changed.

What leaders should do now

The call to action is practical.

First: put energy into the AI business case. Every serious AI initiative should have a compute and energy view, not just a model and vendor view. If the project scales 10x or 100x, what happens to cost, latency, emissions, and capacity?

Second: use real thresholds. A 10 MW workload is a large facility. A 100 MW workload is industrial infrastructure. A 1 GW workload is a regional energy strategy. Treat them differently.

Third: separate high-value intelligence from low-value automation. Not every workflow deserves heavy AI. Use frontier models where judgment, ambiguity, and leverage justify the cost. Use smaller models, retrieval, caching, rules, and process redesign where they are enough.

Fourth: make infrastructure a board-level topic. If AI is strategic, then power supply, data-center capacity, cloud concentration, and sustainability are strategic. CIOs, CTOs, CFOs, COOs, and sustainability leaders need one shared view.

Fifth: build partnerships beyond software. The AI stack now reaches into energy markets, utilities, real estate, cooling, semiconductors, construction, public policy, and eventually maybe space.

The leadership shift

The first AI leadership question was: "What can this technology do?"

The second was: "How does it change work?"

The third is now emerging: "What does it require from the physical world?"

This is where the debate becomes more serious.

AI is not just a software wave. It is a capital investment wave, an energy demand wave, and an infrastructure coordination problem. The limiting factor may not be imagination. It may be megawatts.

Executives should not panic about that. But they should stop treating it as somebody else's problem.

Models matter.

But electricity decides where the models can run. And if the curve continues, the strategic question may become even stranger:

How much intelligence can Earth afford to host?

Sources and further reading
2026-05-17
Hyperscaler 2026 capex hits ~$700B. Free cash flow is the variable that breaks.
What was announced

On February 6, CNBC reported that combined 2026 AI capex commitments across Amazon, Google, Microsoft, and Meta now approach $700 billion. Amazon: roughly $200 billion. Alphabet: up to $185 billion. Microsoft: increase from prior 2025 levels (analyst consensus near $99 billion FY26, ending June). Meta: budgeted $115–135 billion. Approximately 75% of the spend is AI-related — call it $450 billion of AI infrastructure in a single year, up about 36% versus 2025. Free cash flow projections for the same set of companies show meaningful compression; Amazon is forecast to turn negative, with analyst projections of negative free cash flow between $17 billion and $28 billion in 2026.

What it means

Capex of this magnitude rewrites the financial model for the entire frontier compute stack. The hyperscalers are no longer building toward a near-term revenue profile — they are building toward a 5-to-7-year usage curve they believe is coming. That is a different posture than the 2018–2022 capex cycle, which was largely demand-led. This one is conviction-led, and the conviction is asymmetric: if AI compute demand materializes at the projected rate, today’s capex looks conservative; if it lags by even 18 months, the depreciation schedule eats free cash flow at a rate the public markets have not yet priced.

A second-order effect matters more for non-hyperscalers: every CIO planning AI infrastructure in 2026 is now negotiating against a supplier base whose capacity is partially already absorbed by internal hyperscaler workloads. Pricing power for capacity is structurally higher, lead times for premium GPU instances are longer, and the cost-per-token of frontier inference will move on hyperscaler margin compression rather than competition.

Andreas’s view

My read on this: $700 billion is not a number that resolves itself by spreadsheet logic. It resolves itself by which hyperscaler is willing to absorb the cash-flow hit longest. The strategic question inside each company is no longer “should we build” but “which competitor blinks first when the free-cash-flow line turns red on quarterly reporting.” Amazon is closest to that line. Microsoft has the strongest cash position to absorb it. Google sits in between. Meta has the most flexibility because its core ad business is funding the AI infrastructure with the lightest accounting drag.

I don’t think the capex commitment will be revised down materially in 2026. The competitive cost of unilaterally easing off — handing GPU capacity, customer relationships, and the model-training cadence to a competitor — is too high. What will happen instead is creative financing: more debt, more partnerships with sovereign wealth and infrastructure funds, more long-term capacity contracts that move spend off the balance sheet. The capex will continue. The accounting around it will get more interesting.

The way I see it, adjacent businesses should not assume the capacity they need will be available at the price they modeled. My expectation is that premium-tier inference and training capacity will be priced as a scarce resource for the rest of 2026 and most of 2027. Any AI roadmap that depends on flat or declining unit costs over that window has a hidden assumption built in that I think is unlikely to hold.

Three things I’m watching
1. I’ll be watching whether companies move to lock multi-year capacity contracts for premium inference and training now, or wait — because negotiating against scarcity in 2027 will be more expensive than over-committing modestly in 2026.
2. The companies that preserve optionality will be the ones that have stress-tested their AI cost models against a scenario where frontier-tier compute prices are flat or rising for 18 months — and redesigned the workflow, not the budget, when the unit economics broke.
3. Hyperscaler free-cash-flow disclosures over the next four quarters are the leading indicator I’m focused on — they will show whether the capex commitments hold or quietly compress.
References and related signals
- CNBC: Tech AI spending approaches $700 billion in 2026, cash taking big hit (Feb 6, 2026)
- Quartz: Meta beats earnings as 2026 AI capex tops out at $135 billion
- CreditSights: Hyperscaler Capex 2026 Estimates
- theCUBE Research: the $600B annual capex framing and what it implies
- Related signal: Mistral AI raised $830 million in debt financing in late March to build its own data-center capacity in France — sovereign-AI capex is now a parallel curve to the hyperscaler curve, not a substitute for it.
2026-02-08