Right, pull up a stool. Nothing much doing today, is there? Tuesday, market's quiet, despite the 1.4% premarket rise in the Nasdaq for no real reason, and if you've been sat there refreshing your charts hoping for a signal, save your eyes. There isn't one. Not today.

But don't switch off, because the next fortnight is shaping up to be one of those stretches that decides the tone for the rest of the year. Alphabet reports Wednesday evening. Microsoft and Meta follow next Wednesday. Apple and Amazon report the day after that. Between them, these five companies are about to tell the market whether the biggest spending spree in corporate history is actually working — or whether we've all been standing round admiring the emperor's very expensive new data centre.

As a trend trader working top-down — sector, industry, company — this is exactly the kind of week you plan your month around. Not because there's a trade sat waiting for you today, but because what happens over the next ten days sets the character of the tape for weeks after. So let's have a proper look at what's actually going on, why it might spook the market, and — the bit nobody else seems to be talking about — what it means for your pension even if you don't own a single AI share.

The Numbers That Should Make You Sit Up

Between Amazon, Google, Meta and Microsoft, analyst estimates converge on something like $700–725 billion in capital expenditure for 2026 — up roughly three-quarters on last year. Worth being precise about what that figure actually is: it's total capex, not a pure "AI spending" line — it includes ordinary cloud infrastructure, warehouses, networking and other bits alongside the data centres and chips. And it's an analyst aggregation, not a single confirmed number the four companies jointly announced. But even with those caveats, three-quarters more spending in a single year, on top of an already eye-watering base, is the kind of number worth sitting with for a second.

Now, spending money isn't automatically a problem. Companies spend money on infrastructure all the time — it's called capital expenditure, or "capex". The question the market's been quietly chewing over for a couple of years now is simple: is that spending turning into actual profit, and is it doing so fast enough?

Because here's the thing about a top-down approach — sector strength doesn't exist in a vacuum. If the biggest sector in the market (tech, and specifically the AI names within it) can't show it's turning that spend into returns, that weakness doesn't stay neatly parcelled off in one corner of the index. It leaks. And when you're the size AI now is within the likes of the Nasdaq, "leaks" can very quickly become "floods."

Is AI Actually Making Money?

Short answer: some of it is, a lot of it isn't yet, and it depends heavily on which company you mean. Worth being precise here: Alphabet, Microsoft, Amazon, Meta and Apple are established, audited public companies making genuine profits — the uncertainty isn't over whether they're real businesses, it's whether their AI spending specifically is paying for itself yet. The much bigger question marks sit with the private frontier labs — OpenAI and Anthropic — whose detailed financials aren't public yet, and won't be until they actually list.

Allianz Research put a number on the capex-versus-sales problem: something like a 46% gap between how fast AI investment is growing and how fast AI-related sales are growing, cited since in various bits of market commentary. To put that in context, the gap between telecom investment and telecom revenue at the height of the 2001 dot-com bust was about 32%. If that Allianz estimate holds up, we're currently running hotter than the bubble everyone still uses as the textbook example of "companies spent money on infrastructure nobody needed yet."

There's also a tracker that measures what users are actually willing to pay for AI tools — basically, are people happy to keep paying for the token-based AI subscriptions and services, or are they starting to baulk at the cost? That index has fallen about 20% since May. That's not a "the sky is falling" number on its own. Still, it's the kind of thing that makes institutional money nervous, because it hints that pricing power — the ability to charge what you want because people need what you're selling — might be slipping rather than strengthening.

And then there's this cracker from Forrester: fewer than a third of business decision-makers can currently draw a straight line between the money they've spent on AI and any actual improvement in their bottom line. That's not a "does AI work" problem. AI clearly works for plenty of tasks. It's a "can we prove it's worth what we paid for it" problem — and that's the sentence that ends up in every CFO's head come budget-review season.

Quick jargon note: "ROI" just means return on investment — did the money you put in come back to you, with something extra on top, or did it just disappear into the machine. When you hear analysts fretting about AI's ROI, that's all they mean.

So Is It All About OpenAI and Anthropic Now?

You'll have seen the headlines — both OpenAI and Anthropic have confidentially filed paperwork with US regulators, aiming for a public listing later in 2026 or into 2027, potentially valued anywhere from the high hundreds of billions to over a trillion dollars each. Neither has confirmed a firm date. Two companies, from a standing start a few years back, potentially worth more than most entire stock markets.

  • OpenAI burned through roughly $3.7 billion in cash in the first quarter of 2026 alone — about 65% of its $5.7 billion revenue for the same period. Both numbers tripled compared with a year earlier, which cuts both ways: real growth, but the cost of that growth is rising just as fast. The company holds a large cash cushion from recent fundraising, and profitability isn't expected until around the end of the decade.
  • Anthropic's burn rate is reportedly smaller as a share of revenue and shrinking faster, with some projections pointing at an operating profit as soon as mid-2026. Sounds better on the surface. But the "run-rate revenue" figure everyone quotes annualises a single recent month — it isn't the same as revenue actually earned over the past twelve months, and the gap between the two can be substantial. There's also a "circular financing" criticism worth naming as analysis rather than fact: Google and Amazon are simultaneously investors in Anthropic, its computing suppliers, and companies whose own reported profits get a lift when Anthropic's valuation rises. Critics call that a closed loop rather than genuine economic activity. Anthropic's also been in a genuine legal scrap with the US government this year over a "supply-chain risk" designation — a label no other American tech company has had — and while a court has temporarily blocked part of that designation, part of it remains in effect and the case is still working its way through appeal. Separately, Anthropic settled a long-running copyright lawsuit with authors for $1.5 billion, a settlement that received final court approval this week.

Neither company has filed a full set of audited public accounts yet. Until they do, every number quoted by either side — bull or bear — is a projection dressed up as a fact. Treat both the same.

The Bit Nobody's Talking About: Your Pension

Right, here's your "something other blogs don't give you." Forget whether you personally trade AI stocks. This next bit affects you anyway.

Nasdaq changed its listing rules this year to fast-track huge new companies straight into its indexes — as little as 15 trading days after going public for the biggest names, rather than the roughly three-month wait that used to apply. SpaceX has already gone through this fast-track, joining the Nasdaq-100 in July. Worth being precise, though: the S&P 500 looked at doing the same thing and explicitly said no — it kept its existing rules requiring a year of trading history and genuine GAAP profitability, which currently rules out SpaceX, and would rule out OpenAI or Anthropic too, until they can show four straight profitable quarters. So this fast-track story is a Nasdaq-specific one, not a whole-market one.

Here's why it still matters to you even if you've never bought a single AI share: index funds and trackers don't get a choice. If a stock is added to an index, funds that track that index are mechanically required to buy it, in proportion to its size, regardless of price. No judgement call, no "is this good value," just buy because the rulebook says so.

When SpaceX joined the Nasdaq-100, it entered at under 1% weight, funded by small, proportional trims across the other 100-odd holdings — not some single dramatic dumping of other stocks to make room. Money managers at places like MSCI have still flagged the broader dynamic as worth stress-testing portfolios against, and investors like Michael Burry have publicly criticised the whole idea of fast-tracking unprofitable mega caps into passive index funds on principle. But the practical impact on any one Nasdaq-100 tracker from a single new listing is a modest, mechanical nudge — not a bet-the-pension event. The real point worth holding onto is a structural one: if OpenAI and Anthropic both follow SpaceX onto the Nasdaq-100, that's several sub-1% nudges stacking up in the same growth-heavy direction, at a time when the sector's already carrying outsized weight in most growth trackers.

So even the most cautious, "I only hold trackers, I don't gamble" investor has some exposure to this story — just smaller and more mechanical than the headlines make it sound.

Right, So What Actually Happens This Week?

Here's my honest reckoning: nobody knows how this plays out, me included.

I don't think we get a clean "great earnings, market cheers, off to new highs" story this time round. Here's the sequence I think is more likely: earnings come in strong. Revenue up, cloud growth solid, all the numbers a normal company would be thrilled with. But investors, primed by months of "is this bubble about to pop" headlines, look straight past the good numbers and fixate on the bit that scares them — the spending, the ROI gap, the burn rate. Stocks fall anyway, on earnings that, on paper, deserved to send them up.

That's not irrational in the "these people are stupid" sense. It's irrational in the much more common sense: when sentiment has already made its mind up, evidence gets bent to fit the story rather than the other way round. Markets aren't calculating machines matching price to value tick by tick. Most of the time, they're a room full of people reacting to how a headline makes them feel, and adjusting their story to match afterwards.

So the selling continues, perhaps for a few days, perhaps longer, right up until someone in that room asks out loud, "hang on, why are we selling this, the earnings were actually fine?" That's not really a valuation calculation kicking in. It's exhaustion. The bears run out of fresh reasons to keep pressing, and the buyers who've been sitting quietly on the side-lines, waiting for exactly this kind of overreaction, step back in. Not because a spreadsheet told them to. Because the story ran out of fuel.

That, right there, is trading psychology doing more work than any earnings report. Worth sitting with, because it cuts both ways for you as a trader: the crowd's fear can push a genuinely strong company down to a daft price, and the same crowd's short memory can then push it right back up again for no better reason than "well, it stopped falling." Neither move has much to do with what the company is actually worth. Both moves are entirely real, and both will show up on your chart whether you understand why or not.

What This Means for Our Watchlist

If you're a top-down, sentiment-following swing trader like me, here's the practical takeaway. Don't try to predict Wednesday's number. You can't, and neither can anyone selling you a newsletter that claims otherwise. Instead, watch the reaction to the number, not the number itself. A strong print that gets sold hard tells you more about where sentiment currently sits than the actual revenue figure does. That reaction is your signal, not the earnings release.

Keep an eye on whether the weakness stays contained to AI-heavy names, or whether it starts dragging the broader index down with it — that's your tell for whether you're looking at a sector pullback or something with more bite to it. And watch your own head as much as the charts. If you catch yourself buying a dip purely because "it can't keep falling, surely," ask whether that's analysis or just you matching the crowd's mood swing back the other way.

A Word on Risk!

If any of this does turn into a tradeable setup once the dust settles, a couple of basics are worth repeating, because they get forgotten fastest exactly when the market's most exciting.

Position size before you get excited about the story, not after. A "this changes everything" narrative is precisely the moment traders tend to oversize, because the emotion of the story feels like conviction. It isn't. Conviction is knowing where you're wrong and getting out there without an argument with yourself. If AI names do get properly punished, that punishment can arrive fast and with real volatility either side — gap risk on earnings reactions is not a small-print warning, it's a regular Tuesday in this sector at the moment.

Know your stop before you know your target. It's tempting, watching a story this big, to think purely about the upside — new highs, the crowd piling back in, all of that. But the sequence I've laid out above has a version where the selling simply doesn't stop where you expect it to, and the "exhaustion" moment I described takes weeks rather than days to arrive. If you're wrong about timing, your risk management should be doing the deciding, not your hope that the story plays out the way you fancied.

Nothing to Do Today. Plenty to Watch This Week.

Get the kettle on, and I'll see you back here once the numbers actually land.