The Trillion-Dollar Tech Companies Are Afraid. That Should Tell You Something

The billions pouring into AI from the world's largest technology companies look like ambition. Read the underlying dynamics and they look much more like fear.

Audio summary: https://youtu.be/5qkEIeeyIgU

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/the-trillion-dollar-tech-companies?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

The argument I want to make in this article runs against the dominant narrative. The dominant narrative is that Big Tech’s extraordinary AI investments are evidence of confidence — that the world’s most valuable companies see the future clearly and are racing to own it. I want to suggest a different reading. What the investment pattern of Google, Microsoft, Meta, and Amazon reveals, when examined carefully, is not confidence. It is controlled panic. And the distinction matters enormously for every business leader building a strategy around the assumption that these companies remain the permanent infrastructure of the digital economy.

CONTEXT AND BACKGROUND

In 2026, the four hyperscalers — Amazon, Google, Meta, and Microsoft — are on course to spend nearly $700 billion combined on AI infrastructure. Amazon has committed $200 billion. Google has projected between $175 and $185 billion. Meta has outlined between $115 and $135 billion. Microsoft’s annualised run rate points to approximately $145 billion. These are not incremental investments. They represent a roughly 67% spike from the already historic levels reached in 2025, and the largest single-year capital expenditure surge in the history of the technology industry.

At the same time, none of these companies can yet demonstrate a positive return on investment from their AI infrastructure at scale. Amazon is projected to turn free cash flow negative in 2026. Barclays estimates Meta’s free cash flow will drop by almost 90%. Microsoft’s free cash flow is forecast to slide by 28% before recovering in 2027.

A UC Berkeley economist described what is happening plainly. Much of the current AI spending represents defensive insurance rather than profit-seeking investment. Companies feel compelled to build AI capabilities to avoid being left behind, even when immediate returns are unclear. The analogy he drew was to the 1990s browser wars, when Microsoft gave away Internet Explorer for free to prevent Netscape from establishing dominance. The parallel is instructive. Giving away a product for free, or burning cash at a scale that turns your free cash flow negative, is not the behaviour of a company that believes it has a locked competitive position. It is the behaviour of a company that believes it cannot afford not to.

INSIGHT AND ANALYSIS

To understand why these companies are afraid, you need to understand what they built and what is now threatening it. For two decades, the supremacy of Big Tech rested on four pillars: proprietary data, network effects, massive infrastructure, and a monopoly on talent. Each of these is under simultaneous pressure.

The data advantage is shrinking. Google’s search data and Meta’s social graph were once considered irreplaceable assets. Today, AI models trained on the vast expanse of the public internet are matching the performance of models built on closed data sets. The infrastructure advantage is commoditising. The fierce competition between cloud providers is driving down compute costs, making high-level AI capability accessible to any organisation with a credit card. And the talent advantage has fractured — top-tier researchers no longer feel tethered to Big Tech’s compensation structures, choosing instead to lead independent teams.

The most structurally significant threat, however, is open source. In May 2023, an internal Google document was leaked that contained a sentence that has proved more prescient than its authors perhaps intended: “We have no moat, and neither does OpenAI”. The memo argued that open-source AI communities were developing capabilities faster than closed labs, and that the cost of fine-tuning open models was collapsing. At the time, many dismissed it as one engineer’s opinion. It was also correct. DeepSeek’s R1 model, released in January 2025, performed comparably to OpenAI’s most advanced reasoning models at a reported training cost of approximately $5.6 million — a fraction of what American labs had spent on inferior results. It proved that world-class AI did not require Silicon Valley budgets, Silicon Valley infrastructure, or Silicon Valley talent. The assumption that better AI could only be built by those who could afford to build it at scale was invalidated overnight.

I have written previously about the way in which the competitive landscape of AI is shifting and the governance implications that flow from that shift for South African organisations. The structural change now underway is not merely a governance question. It is a strategy question. When open-source models reach a certain threshold of frontier capability, the business model of selling AI access via API begins to collapse. If an organisation can run equivalent intelligence on its own hardware for free, the multi-billion-dollar bets made by Microsoft and Google on proprietary model ecosystems face a genuinely difficult path to return on investment.

What has followed from this realisation is precisely what controlled panic looks like in a corporate context. Non-AI projects are being shuttered. Safety teams have been restructured or disbanded. The “move fast” philosophy has returned at companies that had spent years presenting themselves as responsible stewards of powerful technology. The same week that The New Yorker published its investigation into Sam Altman’s trustworthiness, OpenAI released a 13-page policy paper comparing its vision to the New Deal. The timing is instructive. A company that is genuinely winning does not need to publish a document explaining why it should be trusted with the future. It consolidates. What we are seeing from Big Tech is not consolidation. It is narrative management under pressure.

IMPLICATIONS

For South African business leaders, the instinct when confronted with these investment figures is to read them as signals of where to follow. If Google is spending $185 billion on AI infrastructure, the reasoning goes, the direction is clear, and the smart move is to align with the leaders. This reasoning contains a significant error. The $700 billion being spent by the hyperscalers is not primarily an investment in the future. It is a defensive bet against a present threat. The companies making these investments are not certain they will win. They are certain they cannot afford to lose — and those are very different strategic postures. An organisation that builds its AI strategy around dependency on platforms whose competitive position is genuinely uncertain is not building on solid ground. It is building on the narrative that these companies have spent billions to sustain.

The more productive question for South African executives and boards is what the erosion of Big Tech’s moats actually makes possible. If the barriers to entry are lower than they have been at any point in the past twenty years, if a small team with a superior idea can now realistically challenge a trillion-dollar incumbent, and if the decentralisation of AI capability is genuine rather than rhetorical — then the strategic opportunity for organisations in emerging markets is real and time-limited. Windows of this kind do not stay open indefinitely.

CLOSING TAKEAWAY

The $700 billion being spent by the world’s most valuable technology companies on AI infrastructure in 2026 is one of the most significant economic facts of our time. It is being reported almost universally as a story about ambition, confidence, and the race to own the future. It is more accurately understood as a story about companies that built their dominance on moats they have now admitted they no longer have, spending whatever it takes to rebuild them before the window closes. That is not a reason to dismiss these companies — they have resources, distribution, and institutional capability that remain formidable. But it is a reason to read their investments for what they are: defensive moves by institutions that are genuinely afraid of what is coming, rather than confident bets by institutions that believe they have already won.

Author Bio: Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He served as a working group member contributing recommendations toward South Africa’s national AI strategy, an initiative by the National Advisory Council on Innovation (NACI), the Council for Scientific and Industrial Research (CSIR), the Human Sciences Research Council (HSRC) and the Department of Science and Innovation. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net