BUSINESS REFLECTION: Crossed Wires: What happens to Big Tech now after the midterms?

It is sometimes proposed that swings in stock market prices ultimately reflect the hopes and fears of the citizenry. This is true enough, but also enough of a simplification to be nearly useless. Markets are complex and entangled beasts where company performance, politics, geopolitical risk, wars, natural disasters, economic indicators and the inscrutable calculations of AI-driven trading systems conspire to confound predictions. Peel away the layer of hopes and fears and a bewildering complexity lurks underneath, much of it comprehensible only in retrospect, when everyone suddenly becomes an expert on what was bound to happen.It is against this background that we witnessed last week’s vertigo-inducing sell-off in technology stocks. The casualties were concentrated around the AI ecosystem – chips, data centres, memory manufacturers and the companies financing the infrastructure on which the AI revolution depends. The Nasdaq suffered its worst single-day decline in 19 months, and commentators promptly began dusting off their obituaries for the AI boom.Two factors appear to have combined to produce the carnage – the sweeping Democratic victories in the US midterm elections and growing discomfort with the extraordinary financial commitments supporting the AI-industrial complex. Put the two together, and the market suddenly lost its appetite for these stocks.Transient panicI suggest that the sell-off will prove transient, perhaps even forgettable, although predicting market movements (and especially their timing) is a pastime best left to people who enjoy being publicly wrong (though that hasn’t stopped me before, sometimes to my discomfort).I also believe that describing this as the inevitable popping of the AI bubble is another simplification that obscures more than it reveals. The future of AI and the financial performance of individual companies investing in it are two different matters.Let’s start with politics. The Democrats’ victories, particularly in states such as Texas, have been interpreted as at least partially a referendum on the unbridled power of GOP-supporting tech bros and their billion-dollar bank accounts, certainty and braggadocio. There is a growing narrative that tougher environmental regulation, opposition from local communities and political hostility towards Big Tech will substantially slow the industry’s headlong expansion into AI, including (most visibly) AI data centre development.There is some truth here, but it is hardly the whole story. The public backlash against data centres predates the election and is not confined to Democrats. Republican communities have been complaining about electricity prices, water consumption and the distinctly underwhelming employment benefits of these enormous facilities for years. Residents have begun asking why they should tolerate the inconvenience while technology companies collect the profits. It is a reasonable question, irrespective of political affiliation.But there are several reasons to believe that data-centre development will continue, even under a more demanding regulatory regime. The first is geography. Computing is unusually mobile. Unlike a steel mill or car factory, a data centre does not necessarily have to be located near its customers. AI model training, in particular, can be conducted wherever electricity, connectivity and suitable infrastructure are available. This creates opportunities beyond the United States (Mexico, for instance), including countries with abundant renewable energy and relatively accommodating regulatory environments. There are even promising experiments to move computing into the oceans, or eventually into orbit, although the latter remains some distance from commercial reality.Moving data centres is not frictionless, of course. Latency matters for some applications, electricity grids are expensive to develop, and governments increasingly regard computing infrastructure as strategically important. Nevertheless, the technology industry has considerably more geographical flexibility than many traditional industries. If one jurisdiction becomes prohibitively expensive or restrictive, another may become more attractive. The bits and bytes being processed and moved around are, after all, utterly indifferent to the jurisdiction in which they were spawned.Putting the capital in capitalismThe second reason is the transactional nature of capitalism. Suppose a developer approaches a community with an offer – permit us to build a data centre over the hill, and we will guarantee that your electricity bills will not increase. We will pay for the necessary grid upgrades, contribute substantial local taxes to assist your schools, hospitals and other services, and invest in water recycling. Suddenly the conversation changes. Residents may still ask questions about noise, environmental damage or the intrusion itself, but the discussion has moved from outright opposition to a negotiation. How much compensation? What guarantees? Who pays if electricity costs rise? These become questions of money, leverage and enforceable obligations rather than an outright refusal to engage. In other words, everything has a price.At the federal level, there is likely to be a move to rein in the current administration’s somewhat laissez-faire attitude towards private AI development. But the truth is that AI has not been without oversight. Federal scrutiny in the US has expanded, with leading frontier AI companies voluntarily providing government researchers with early access to their most powerful models for safety and national-security testing. A June 2026 executive order formalised a framework for such evaluations up to 30 days before release, although it stopped short of making them compulsory. They may become compulsory under a Democratic Congress, but that will make little difference to the march of AI progress.There are, to be sure, people who would hobble the industry. Bernie Sanders and Alexandria Ocasio-Cortez introduced an AI Data Center Moratorium Act in March, and gubernatorial candidates in at least a dozen states backed moratoriums. But the centrists who hold sway over the party’s louder wings know two things. AI is the most transformative technology in history, with a growing likelihood of genuinely utopian outcomes in areas like medicine, education, drug discovery and energy. And no US legislator wants to be remembered as the one who lost the race to China. Scrutiny will sharpen, but the brakes will not be slammed on.Most importantly, even with increased Democratic-led oversight and regulation of AI, there is no structural change in the demand for AI, which is still growing by double- and triple-digit percentages annually across almost any sector of human endeavour you care to look at. Compare this to the financial crash of 2008. In that case, there was a structural collapse in demand in the wake of incorrectly rated and priced mortgage loans. An unwinding of that debt was inevitable – the mathematics of finance does not bend. Those loans had to be repaid or forgiven; balance sheets are immovable.In contrast to the subprime crisis, demand for AI was the same the day after the election as the day before. By the spring, Microsoft’s AI business had passed a $37-billion annual run rate, up 123% year on year, and Alphabet had lifted its 2026 capex guidance to between $180-billion and $190-billion. There is scarcity of just about everything up and down the AI supply chain, and scarcity means rising prices. No election changes that, and markets must obey it.The hidden debt crisisWhich brings us to the debt. The headline numbers are vertiginous. Morgan Stanley’s research puts global data-centre capex at about $2.9-trillion between 2025 and 2028, and Moody's reckons at least $3-trillion over five years. The bank expects AI-related debt issuance of nearly $570-billion this year alone.Can it be repaid? The subprime borrowers before 2008 were households that could not service teaser-rate mortgages that once looked too good to be true (they were). The bulk of today’s AI borrowers are investment-grade hyperscalers with enormous operating cash flows, selling a product that is already earning revenue. I am aware that “this time is different” are the four most expensive words in finance. But I am saying it anyway, because the borrowers are indeed different.Of course, the recovery from last week’s sell-off will be spiky. There will be dead-cat bounces and trading-bot dip-buyers and sneaky arbitragers and leverage sharks. But there is no stopping the train, and cautious professional investors probably know that, and will wait out the current nervous shudder.Big Tech’s most vocal barkers and backers – Musk, Thiel, Karp, Andreessen, Altman, et al – are not about to retreat and wait for a Republican recovery. The money that they spent on their GOP allies may have been lost, but new money will be directed at eager takers outside that tent, either in the Democratic Party or beyond the US borders. And markets care far less about where donations land than about whether chips ship and tokens sell. The next real test arrives when Nvidia reports on 17 November. By then, I suspect, the market will have largely forgotten last week.AI companies are providing a product with endless upside for humanity (yes, there will be downsides too). A cure for cancer, dramatically improved education for children, cheaper and safer modes of transport and higher business productivity are already (or will be) on the shelves – there will be buyers aplenty, and these markets will recover quickly.So what happens to Big Tech now? Not much. They will just spend some more money to work around any regulations they don’t like. Supply eventually obeys demand. DMSteven Boykey Sidley is a visiting professor at the Centre for Graduate Studies at Rhodes University, a partner at Bridge Capital and a columnist-at-large at Daily Maverick, where he writes the weekly Crossed Wires column. His new book, Checkmate: How Shoprite Checkers rewrote the rules and won, is published by Jonathan Ball.