Corporate profits are skyrocketing as the AI boom continues, while workers continue to lose out.
Michael Roberts is an Economist in the City of London and a prolific blogger.
Cross-posted from Michael Roberts’ blog
As the central bankers of the world met at Jackson Hole, Wyoming for their annual get together to discuss issues affecting the world financial sector and monetary policy, US statistics for corporate profits for the second quarter of 2026 were released. The figures were truly startling. US corporate profits surged by $401 billion in the second quarter of 2026. This pushed total corporate profits to a historic high of $4.3 trillion, or up nearly 23% since Q2 2025.
and pushed gross profit margins (ratio of profits to sales) to 19.4%—the highest recorded in the US since the 1940s!
Both capitalists and Marxists agree: profits matter; they are the driving force for investment in a capitalist economy. When profits are rising, investment follows and eventually employment even if the investment is labour-saving like AI supposedly is. So these figures tell you that a recession in the US is not yet on the agenda.
What is happening here to deliver such a revival in US corporate profits? It appears to be a combination of aggressive corporate price hikes following the end of the pandemic, massive AI infrastructure spending leading to huge profits for those building the datacenters, making chips and other facilities for the AI boom; and Trump’s corporate profit tax cuts. These tax cuts have added nearly $50bn to corporate profits after tax for the top 100 US companies, while hardware and chip makers (the AI “sellers”) have experienced explosive earnings, with tech sector earnings rising more than 65% year-over-year.
But a key factor in driving up profits has been the general suppression of workers’ wages. While pre-tax profits as a share of national income hit 18%, the highest share since the aftermath of the second world war, employees’ share from wages and benefits fell to 60 per cent, the lowest level since the 1950s. Inflation has outpaced wage growth, causing real hourly earnings to fall by 0.2 per cent in July versus a year earlier.
Chief executives at America’s largest low-wage employers have seen their pay rise by 41 per cent between 2019 and 2025, while the average worker took home only 21 per cent more — that’s below the 26 per cent increase in prices over the period. As the FT put it: “Workers have been gradually losing sway in corporate America since the early 1980s as union membership has declined and companies increasingly outsource jobs to external contractors.” But the decline in labour’s share of income has gained pace since the end of the COVID pandemic and especially over the past 12 months.
The boom in profits is still concentrated in the Magnificent Seven tech companies. They have raised revenue by 36%, but profit even more by 67%, according to work by Brian Green. “This increase in the mass of profits of two thirds in just one year which goes a long way to explain why Wall Street has been so buoyant.”
But there are some caveats to these headline figures. It seems that the tech giants have heavily padded their headline profit figures via “other income” accounting items, capturing massive unrealized investment gains from their stakes in private startups and AI ventures. The big tech companies booked more than $160bn in gains from investments in owning the shares of other AI companies last quarter. Alphabet, Amazon, Nvidia and Microsoft all recorded significant boosts to pre-tax profits through valuation gains on equity stakes, including investments in OpenAI, Anthropic and SpaceX. Alphabet recorded $97.9bn in “other income” in the three months to June 30, while Amazon reported $53.4bn. Nvidia recorded a further $7.7bn in the three months to the end of July.
Moreover, the massive investment into AI models and data centres by the Big Tech ‘hypescalers’ looks secure only because the AI companies like OpenAI and Anthropic have a very dubious way of calculating their revenue from AI users, called annual recurring revenue (ARR). They take a recent month’s revenue and multiply it by 12 to set an expectation for the year to come. As Ed Zitron, the top analyst on AI business models, put it “ARR can mean everything from “[actual month] x 12” to“[30 day period of revenue] x 12” and in most cases it’s a number that doesn’t factor in churn. If you use ARR, you’re essentially taking one month and treating it as representative of the entire calendar year, when it isn’t.” Subscription revenues aren’t enough to justify a more than 600 per cent increase in annualised revenue in less than a year. That assumes that non-subscription revenue will stay high even as the industry enters a price war and is struggling to get customers to pay for the most advanced models.
Moreover, the AI companies and the hyperscalers are increasingly raising credit to pay for their investments. Companies had already tapped debt markets for $217 billion last year, according to Morgan Stanley. By mid-August, they had already smashed past this total, with $445 billion of debt issued and the overall debt splurge will reach nearly $600 billion for 2026. That is more than the combined 2026 budgets for the US Departments of Justice, Transportation and Education. Future capital expenditure plans amount to roughly 3% of US GDP a year up to 2030, more than the annual real GDP growth of the whole economy.
In the dot-com bubble and bust, the US stock market lost almost half its value between the 2000 peak and the 2003 trough, but the economy suffered only a mild and brief setback. But this time, technology companies are sucking up all their cash and more to invest in the AI boom. Google’s free cash flows went negative for the first time ever in the second quarter of 2026, after cumulatively generating almost $600 billion since the listing in 2004.
Then there is the issue of more obscure liabilities. The hyperscalers are increasingly keeping debt off their books by structuring their data center investments as ‘lease commitments’. Meta’s Hyperion data center project in Louisiana is not supposedly financed by debt. The company secured most of the financing money by promising to ‘rent’ the data center itself for 20 years. As a result, the $27.3 billion loan appears nowhere on Meta’s balance sheet. Goldman Sachs analysts recently tallied $1.5 trillion in these ‘lease commitments’ by the AI hyperscalers: “this treatment can understate leverage and future liquidity needs as these obligations are eventually recognized and contractual payments come due.” And then there is another estimated $1.5 trillion of “purchase commitments” —promises to buy chips and electricity. Those promises don’t appear on balance sheets either. So while corporate profits have rocketed up, it is being partly driven by the huge financing of AI by hyperscalers in a circular financing model.
Nvidia is increasingly using its own balance sheet to keep the AI boom running. Nvidia is not only selling the GPUs powering the AI buildout, it is also helping finance the infrastructure and customers buying them, including nearly $50 billion invested in AI labs and a planned $105 billion backstop for the Ohio data-center project tied to OpenAI and SB Energy. Nvidia is also working with Wall Street firms on up to $500 billion of financing for AI chip purchases, while OpenAI could purchase roughly $350 billion of Nvidia chips for the full Ohio buildout.
The strategy can supercharge Nvidia’s revenue as long as AI demand keeps accelerating, but it also increases Nvidia’s exposure if AI spending starts to slow. If AI labs and data-center developers cannot fund the buildout themselves, Nvidia may increasingly have to support the customers generating demand for its own chips. The bigger the commitments become, the more painful a demand slowdown could be, because Nvidia would be exposed not only through weaker chip sales, but also through the financing and guarantees supporting the AI boom.
Meanwhile AI model usage prices are collapsing as competition from Chinese models intensifies and yet the cost of GPUs, memory, servers and power remains extremely high. This is creating a growing gap between what customers pay and what it costs to build the infrastructure behind AI.
Cheaper AI could drive massive growth in usage, but the bigger question is whether that growth translates into real profits. BCA reckons that AI companies will need to generate $10 trillion a year in revenue just to justify the capex being spent, roughly equivalent to annual global spending on food or healthcare! AI may be getting dramatically cheaper to use, but if the payoff fails to catch up with the cost of building it, investors could soon discover that the AI boom is generating plenty of revenue without nearly enough profit.
Indeed, in Marxist terms, this huge rise in capital investment leads to a rise in the organic composition of capital (the ratio of investment in means of production to the cost of employing workers). That would drive down profitability. So far this has been counteracted by the rise in the rate of exploitation of workers. So corporate profitability (profit per capital invested) is up, if not nearly as much as corporate profits. Since the pandemic, US corporate profitability has risen over 20%, but is still well below the level achieved in the 2010-14 recovery from the Great Recession (but note Q2 2026 profitability data are not available yet). And see Brian Green’s estimates.
Source: Federal Reserve, author calculation.
So the jury is still out on the success of the AI boom; and the likelihood of an AI stock market bust remains high. Nevertheless, optismism prevails, not only among AI companies; hyperscalers and stock market investors, but also among the international powers that be. Indeed, according to Kristalina Georgieva, the head of the IMF, the boom in AI investment is spreading from the US to power the world economy, and is likely to boost global growth this year. “What started out as a US phenomenon with AI is now becoming a growth engine for the global economy, with other countries ramping up construction of data centres and other infrastructure.” Other countries were now “plugged into the supply chain” and exporting AI hardware to the US.
It’s true that global corporate profits (my estimates) have risen sharply since the beginning of 2025 as the AI boom has gathered strength.
But US real GDP growth slowed to 2.1% yoy in Q2 from 2.7% yoy in Q1 and inflation is rising at over 4% a year, with forecasts suggesting it could reach 5% by year end if the Iran war is not resolved. If the Federal Reserve decides to raise its policy interest rate to try and curb inflation, that could trigger a sharp drop in borrowing and a stock market ‘correction’.
Let me end by quoting Ed Zitron on the wider AI picture. “Everyone seems so obsessed with sinking billions of dollars into the theoretical chance that machine learning might be able to replace human beings, and that more money makes it “smarter” and “better” at tasks. Surely investing real money in actual workers — making their lives better, improving their working conditions, teaching them new things, sharpening their existing skills, rewarding them for their hard work, and so on — would have better effects than fastballing hundreds of billions of dollars into a machine that does an impression of work?”
Be the first to comment