SUPPORTED BY

Hi Reader,

Happy Friday! Well, it's not coming home. I always said beware of home bias.

In-depth: Can AI turn a profit in a world of open-source Chinese models?
Newsbites: Slow growth, record bank profits, and IBM's awful day.
PensionCraft News: Will the UK state pension still exist? Is America still exceptional? And our latest tracker updates.

IN-DEPTH

Cash Landing: Will AI Make Money?

Do you like graphs? Of course you do. You wouldn't be here if you didn't. How’s this one for a snapshot of the stock market in 2026?

The hyperscalers — Amazon, Alphabet, Meta, Microsoft, Oracle — will collectively spend more than $725bn on AI infrastructure this year. That’s 77% more than last year. And next year's capex is expected to clear $1 trillion. No one has ever scaled hype like this.

The bill is about to arrive. Analyst forecasts suggest their combined free cash flow (FCF) will collapse to just $4bn in the third quarter — the lowest since 2014, when their revenues were a seventh the size and "the cloud" was still mostly weather. Alphabet's first-quarter free cash flow was already down 47%, and Amazon's has fallen 95% over the past year. Morgan Stanley expects Amazon’s FCF to sink to minus $17bn by December. At which point it is not free, or even cash flow.

Hence a sudden enthusiasm for other people's money. AI-related bond issuance hit $236bn by the end of May, running at four times last year's pace. A further $662bn of data-centre leases are stored off balance sheet, where all the best liabilities live. And Google has just completed the largest equity raise in history.

But the money isn’t disappearing into thin air. It’s building the greatest air-conditioned sheds the world has ever seen, and lining the pockets of semiconductor companies. As the graph makes clear, one company's capex is another company's revenue.

Nvidia's FCF surged from $3.8bn in 2023 to $97bn last year. In the first quarter of this year, it reached $48.6bn — half of last year's total, in just thirteen weeks. At this point its main operational constraint is how fast people can pour concrete. Similarly, Broadcom converts 46 cents of every dollar of revenue into cash. And TSMC, which does the actual manufacturing while everyone else holds press conferences, threw off $32bn last year.

Up until recently, the Mag 7’s valuations rested on the fact these were capital-light machines, converting customers into cash without needing to buy much of anything. That is no longer true. But for now, they remain priced something like software stocks while spending like Victorian railways — a genre of company remembered mostly for awe-inspiring viaducts and irate creditors.

None of this necessarily spells doom. Amazon ran near-zero FCF for years while it built AWS, and shareholders did just fine (if they managed to sit on their hands). Perhaps it will turn out to be exactly what it looks like — companies spending to meet insatiable demand. AWS just posted its fastest growth in fifteen quarters, Microsoft's AI revenue is running at $37bn a year and growing 123%, and Google Cloud's backlog stands at $462bn. The trouble is, in markets, things are rarely exactly what they look like.

So what are the big risks that could deflate the AI “bubble”?

Chinese open-weight models. The biggest threat to AI profitability is that intelligence becomes a commodity, like wheat, or podcasts.

This week, Moonshot AI announced Kimi K3, an AI model with capabilities approaching those of cutting-edge US labs such as Anthropic and OpenAI.

We’ve been here before. In January 2025, Deepseek triggered a global tech selloff after releasing its R1 model, which was competitive with the best US models at the time, at a fraction of the cost. Nvidia lost $593bn in value in a single day. The impact was short-lived as the US models pushed ahead once more, and customers continued to sign up for paid accounts.

The assumption has been that Chinese open-weight models are roughly six months behind US ones in terms of performance. That was always a fragile lead, and the gap might be closing. Why pay a small fortune for US AI if China can give you more-or-less the same thing at a much cheaper price?

However, the developers of Kimi K3 have made clear it isn’t quite up to the level of the frontier US models.

“Despite being a highly competitive model overall, K3 nonetheless exhibits a noticeable gap in user experience compared with Claude Fable 5 and GPT 5.6 Sol.” — Moonshot AI

There's also no guarantee the free stuff keeps coming. Washington already treats its frontier labs as strategic assets, tightening oversight on cyber and national security grounds. Beijing needs even less of an excuse. If Chinese models keep closing the gap, how long before the Communist Party decides that handing them to the world, unrestricted and free of charge, is a strange use of a strategic advantage?

Technological plateau. The models themselves could hit an awkward ceiling where they have read everything humanity has ever written and, like many extremely well-read individuals, find it difficult to monetise.

Bloomberg reports that Google has delayed the release of its latest flagship Gemini model due to underwhelming performance, particularly in coding. But one delayed launch is not a plateau. AI models are now regularly providing novel proofs for maths problems that have remained unsolved for decades. If there is to be a slowdown, we haven’t seen it yet.

Capex treadmill. The opposite risk is that there is no hard ceiling to AI capabilities… but it’s never profitable. Perhaps staying at the frontier requires spending faster than revenue can arrive. In a world of rapidly improving AI, each generation of model could go stale before the chips that trained it are paid off.

As it stands, OpenAI loses around $21bn on $13bn of revenue. I guess it’s still a nonprofit after all.

Regulatory hammer. Governments get scared of the social, economic and security implications, and try to slow the whole thing down. Brussels has already passed an AI Act it doesn't quite know how to enforce, and Washington has shown it can shut down frontier models with as little as 90 minutes' notice.

Human extinction. Or governments do nothing, the machines wake up, try to optimise the planet, and decide to eliminate the world’s biggest underperforming asset… us. Maybe our investments keep rising regardless. Maybe the robots keep paying dividends into accounts nobody will ever check and the claim that ‘the best investors are dead investors’ is at last confirmed at scale.

Trading 212 offers the cheapest way to invest in the markets and build wealth with zero commission and the lowest FX fee on the market, along with multi-currency accounts;
You can very easily automate your investing with Pies;
We have launched an attractive Cash ISA where interest is calculated daily and paid monthly plus it’s a flexible Cash ISA (as is the S&S ISA);

Trading 212 is offering free fractional shares worth up to ‎£⁠100 for new customers: create and verify a Trading 212 account, make a minimum deposit of £1 and use our promo code "RAMIN".

Sponsored Link. Terms apply: trading212.com/join/ramin

When investing, your capital is at risk. Past performance doesn’t guarantee future results. Other fees may apply. Interest applies on the uninvested cash in your account. Free shares can be fractional. Terms apply. T212 Cards are issued by Paynetics which provide all payment services. T212 provides customer support and user interface.

NEWSBITEs

The UK economy eked out 0.1% growth in May. Services output was up 0.3% month-on-month, but was partly offset by a 0.5% decline in production and a 0.8% fall in construction. Meanwhile, sterling hit a one-year high on reports incoming PM Andy Burnham will overlook Ed Miliband to appoint centrist Shabana Mahmood as Chancellor. One day we’ll get our promised "chaos with Ed Miliband".

US inflation slowed to 3.5% in June as petrol prices tumbled. Headline prices fell outright month-on-month, while core inflation eased to 2.6% year-on-year. Markets pushed rate-hike bets back to October, though Fed chair Kevin Warsh refused to declare "mission accomplished". Not when the Strait of Hormuz is still in dire straits.

China's economy grew 4.3% in the second quarter, its slowest pace in three years. A consumer spending slump and unresolved property crisis offset an AI-driven export boom, fuelling expectations that Beijing will unleash further stimulus this year. Better stimulate than never.

Wall Street's biggest banks posted record trading revenues. Volatility from the AI boom and Iran war kept clients trading heavily, with Goldman making $7.4bn from equities alone. Morgan Stanley ($6.3bn) and JPMorgan ($6bn) weren’t far behind. Underwriting boomed too as SpaceX's $86bn listing boosted investment banks’ bottom lines further. Time to counterparty like it’s 1999.

IBM shares plunged 25% on their worst day since 1968. Customers diverted IT budgets to AI servers and memory chips, leaving IBM’s infrastructure sales down 7% and software growth at 5%, below estimates. "We faltered," CEO Arvind Krishna conceded, as Workday, ServiceNow and Salesforce also fell. SaaSpocalypse Now?