colossal

Several Big Tech companies have reported earnings in recent weeks and provided estimates for their spending in 2026, along with leading analysts’ projections.

ADVERTISEMENT

ADVERTISEMENT

The data point that seems to have caught Wall Street’s attention the most is the estimated capital expenditure (CapEx) for this year, which collectively represents an investment of over $700bn (€590bn) in AI infrastructure.

That is more than the entire nominal GDP of Sweden for 2025, one of Europe’s largest economies, as per IMF estimates.

Global chip sales are also projected to reach $1tn (€842bn) for the first time this year, according to the US Semiconductor Industry Association.

In addition, major banks and consulting firms, such as JPMorgan Chase and McKinsey, project that total AI CapEx will surpass $5tn (€4.2tn) by 2030, driven by “astronomical demand” for compute.

CapEx refers to funds a company spends to build, improve or maintain long-term assets like property, equipment and technology. These investments are meant to boost the firm’s capacity and efficiency over several years.

The expenditure is also not fully deducted in the same year. CapEx costs are capitalised on the balance sheet and gradually expensed through depreciation, representing a key indicator of how a company is investing in its future growth and operational strength.

The leap this year confirms a definitive pivot that began in 2025, when Big Tech is estimated to have spent around $400bn (€337bn) on AI CapEx.

As Nvidia founder and CEO Jensen Huang has repeatedly stated, including at the World Economic Forum in Davos last month, we are witnessing “the largest infrastructure build-out in human history”.

Hyperscalers bet the house

At the top of the spending hierarchy for 2026 sits Amazon, which alone is guiding to invest a mammoth $200bn (€170bn).

To put the number into perspective, the company’s individual AI CapEx guidance for this year surpasses the combined nominal GDP of the three Baltic countries in 2025, according to IMF projections.

Alphabet, Google’s parent company, follows with $185bn (€155bn), while Microsoft and Meta are set to deploy $145bn (€122bn) and $135bn (€113bn) respectively.

Oracle also raised its 2026 CapEx to $50bn (€42.1bn), nearly $15bn (€12.6bn) above earlier estimates.

Additionally, Tesla projects double the spending with almost $20bn (€16.8bn), primarily to scale its robotaxi fleet and advance the development of the Optimus humanoid robot.

Another of Elon Musk’s companies, xAI, will also spend at least $30bn (€25.2bn) in 2026.

A new $20bn (€16.8bn) data centre named MACROHARDRR will be built in Mississippi, which Governor Tate Reeves stated is “the largest private sector investment in the state’s history”.

xAI will also expand the so-called Colossus, a cluster of data centres in Tennessee that has been described by Musk as the world’s largest AI supercomputer.

Furthermore, the company was acquired by SpaceX in an all-stock transaction at the start of this month.

The merger valued SpaceX at $1tn (€842bn) and xAI at $250bn (€210bn), creating an entity worth $1.25tn (€1.05tn), reputedly the largest private company by valuation in history.

There are also reports that SpaceX intends to IPO sometime this year, with Morgan Stanley allegedly in talks to manage the offering that now includes exposure to xAI.

Elon Musk stated that the goal is to build an “integrated innovation engine” combining AI, rockets and satellite internet, with long-term plans that include space-based data centres powered by solar energy.

Conversely, Apple continues to lag in spending with “only” a projected $13bn (€10.9bn).

However, the company announced a multi-year partnership with Google last month to integrate Gemini AI models into the next generation of Apple Intelligence.

Specifically, the collaboration will focus on overhauling Siri and enhancing on-device AI features. Therefore, one could say that Apple is outsourcing a lot of the investment it needs to be competitive on AI development.

As for Nvidia, it will report earnings and release projections on 25 February.

The company is primarily in the business of selling AI chips, and is expected to get the lion’s share of the Big Tech’s spending. Particularly, for the build-out of data centres.

In last August’s earnings call, CEO Jensen Huang estimated a cost per gigawatt of data centre capacity between $50bn (€42.1bn) and $60bn (€50.5bn), with about $35bn (€29.5bn) of each investment going towards Nvidia hardware.

The great capital rotation

Wall Street has had mixed feelings about the enormous spending Big Tech companies have planned for 2026.

On the one hand, investors understand the necessity and urgency of developing a competitive edge in the artificial intelligence age.

On the other, the sheer scale of the spending has also spooked some shareholders. The market’s tolerance hinges on demonstrable ROI from this year onwards, as the investments are also increasingly financed with massive debt raises.

Morgan Stanley estimates that hyperscalers will borrow around $400bn (€337bn) in 2026, more than double the $165bn (€139bn) that was loaned out in 2025.

This surge could push the total issuance of high-grade US corporate bonds to a record $2.25tn (€1.9tn) this year.

Currently, projected AI revenue for 2026 is nowhere near matching the spending, and there are valid concerns. For instance, the possibility of hardware rapidly depreciating due to innovation, and other high operational costs such as energy usage.

It can be confidently stated that the numbers have a heavy reliance on future success.

As Google CEO Sundar Pichai acknowledged this month, there are “elements of irrationality in the current spending pace”.

Back in November, Alex Haissl, an analyst at Rothschild & Co, became a dissenting voice as he downgraded ratings for Amazon and Microsoft.

In a note to clients, the analyst wrote “investors are valuing Amazon and Microsoft’s CapEx plans as if cloud-1.0 economics still applied”, referring to the low-cost structure of cloud-based services that allowed Big Tech firms to scale in the last two decades.

However, the analyst added “there are a few problems that suggest the AI boom likely won’t play out in the same way, and it is probably far more costly than investors realise”.

This view is also shared by Michael Burry, who is best known for being among the first investors to predict and profit from the subprime mortgage crisis in 2008. Burry has argued that the current AI boom is a potential bubble pointing to unsustainable CapEx.

Big Tech’s AI race is funded by a tremendous amount of leverage. Whether this strategy will pay off, and which companies will be the winners and the losers, only time will tell.

At the moment, Nvidia certainly seems to be a great beneficiary. Moreover, Apple has a distinct approach by increasing third party reliance, through a partnership with Google, instead of massively scaling their spending. It is a different trade-off.

Europe’s industrial deficit

Amid all this spending, urgent questions have also been raised about Europe’s ability to compete in a race that has become a battle of balance sheets.

For the European Union, the transatlantic contrast is sobering. While American firms are mobilising nearly €600bn in a single year, the EU’s coordinated efforts do not even match the financial firepower of the lowest spender among the US tech titans.

Brussels has attempted to rally with the AI Factories initiative, and the AI Continent Action Plan launched last April, which aim to mobilise public-private investments.

However, the numbers tell a stark story. Total European spending on sovereign cloud data infrastructure is forecast to reach just €10.6bn in 2026.

While this is a respectable 83% increase year-on-year, it remains a rounding error compared to the US AI build-out.

Last year, at the time when the initiatives mentioned were being discussed, the CEO of the French unicorn Mistral AI, Arthur Mensch, stated that “US companies are building the equivalent of a new Apollo program every year”.

Mensch also added that “Europe is building excellent regulation with the AI Act, but you cannot regulate your way to computing supremacy”.

Mistral represents one of the only flickers of European resistance in the AI race. The French company is employing the same strategy as most of Big Tech and aggressively expanding its physical footprint.

In September 2025, Mistral AI raised a €1.7bn Series C at a valuation of almost €12bn, with the Dutch semiconductor giant ASML leading the round by singly investing €1.3bn.

During the World Economic Forum in Davos last month, Mistral’s CEO confirmed a €1bn CapEx plan for 2026.

Just last week, the company also announced a major €1.2bn investment to build a data centre in Borlänge, Sweden.

In a partnership with the Swedish operator, EcoDataCenter, the facility will be designed to offer “sovereign compute” compliant with the EU’s strict data standards, and leveraging Sweden’s abundant green energy.

Set to open in 2027, this data centre will provide the high-performance computing required to train and deploy Mistral’s next-generation AI models.

This is an important move for the company, as it is the first infrastructure project outside France, and it is also a core venture for European data sovereignty.

Meanwhile, US tech titans are attempting to placate European regulators by offering “sovereign-light” solutions. Several Big Tech projects have been rolled out for “localised cloud zones”, for example in Germany and Portugal, promising data residency.

However, critics argue these remain technically dependent on US parent companies, leaving the European industry vulnerable to the whims of the American economy and foreign policy.

As 2026 unfolds, the stakes are clear. The US is betting the house, and its credit rating, on AI dominance.

Europe, cautious and capital-constrained, is hoping that targeted investments and regulation will be enough to carve out a sovereign niche in a world increasingly run on American technology.