TIDE Capital: We’re Still Bullish on AI. But the Reasons Have Changed.
The Market Just Switched Scripts
The market right now is somewhere between excited and nervous. SpaceX just closed a $75 billion mega-IPO. OpenAI and Anthropic are both reportedly preparing their own. At the same time, Alphabet is planning an $80 billion equity raise, and Meta is lining up fresh financing of its own.
Honestly, when this many giants reach into the market’s pocket at the same time, it’s hard for anyone to stay calm. But reading this wave as “AI has peaked” is a little too easy. It looks more like the AI story has just turned to its next act.
For the past two years, the market was buying demand and imagination — the question was whether AI was actually for real. By 2026, the question has changed: how long can this level of investment intensity hold up?
As Tidal Investment founder Jasper Wu puts it,
What markets see is always the fast variables. What actually decides the direction of a cycle is usually the slow ones.
Standing in mid-2026, we’re still bullish on the AI value chain. But our bullishness today can’t rest on imagination alone. Two years ago, you could pitch AI by talking about models and AGI. Try that today, and the market won’t necessarily buy it.
The Money Is Still Going In — And Going In Harder
How do you tell whether a cycle has run its course? Look at whether the people writing the checks are still writing them. Flip through the books of the five biggest cloud companies, and the answer is pretty clear.
Alphabet’s 2025 CapEx was $90 billion, with 2026 guidance raised to $180 billion. Amazon’s 2025 CapEx was $130 billion, with 2026 guidance raised to $200 billion. The other three are moving the same direction: Meta’s 2026 guide is up to $140 billion, Microsoft’s to $190 billion, and Oracle’s FY26 is approaching $60 billion.
Laid out like this, the numbers are a little intimidating. The hardest thing about these internet giants used to be how strong their cash flow was and how much cash they sat on. Now, standing in front of AI, even they are more aggressively reaching into the capital markets. Beyond that $80 billion equity raise, Alphabet has also issued meaningful debt over the past year. AI infrastructure has gotten so large that even the best-capitalized companies on the planet are rethinking their capital structure.
The money is still going in. That’s not in dispute. The question is: how long can it keep going in at this pace?
Why This Cycle Doesn’t Stop That Easily
What does everyone worry about most? They worry CapEx will peak. They worry this will play out like past tech hardware cycles — two or three years of frenzied buying, followed by a long digestion phase. Servers, smartphones, PCs — a lot of hardware cycles work this way: demand picks up, capacity expands, inventory piles up, and the moment demand slows, the whole chain gets de-rated together.
In past cycles, that worry was fair. This round of AI CapEx probably isn’t that simple.
First, the money is going into too many different places. On the surface it’s all called “CapEx,” but if you actually break it apart, it’s not one thing: compute, memory, networking, power — each layer has its own expansion rhythm and its own bottleneck. And engineering projects have a particular nature: once you’ve started, pulling out halfway is more expensive than just gritting your teeth and finishing.
More importantly, the bottleneck is moving from chips down into the more physical layers. Chip shortages can be solved by expanding fab capacity. But power, transformers, high-density racks — those can’t be expanded that fast. Just getting connected to the grid often means waiting in line for years.
And CapEx stopped being just GPUs a long time ago. The signal from the supply chain is clear: Eaton, which makes power distribution equipment, saw data center orders grow 240% year-over-year in Q1 2026.
Transformers, UPS systems, liquid cooling, thermal management, rack integration — these only show up in volume when the hyperscalers have committed to actually building campuses. When all of these orders surge together, it means there’s real construction progress underneath the CapEx cycle.
Put it all together, and you can see why this round doesn’t stop that easily.
What the Market Is Actually Worried About
Being bullish doesn’t mean ignoring the two concerns that are very much on the market’s mind right now.
Concern #1: CapEx is growing faster than revenue. Will the ROI show up?
For the Big Five clouds, 2025 CapEx growth outran revenue growth across the board. Alphabet’s depreciation rose from $15.3 billion in 2024 to $21.1 billion in 2025 — a 38% jump in one year, and it’s now real, on the income statement. Amazon basically said the quiet part out loud in its filing: free cash flow is declining because AI investment is pushing PP&E up.
There’s a popular line in markets right now: once CapEx growth outruns revenue growth, ROI has peaked. The line isn’t wrong, but applied to the cloud business it’s a bit lazy. AWS, Azure, and GCP all went through long stretches in the early 2010s when CapEx blew past revenue, and they ultimately monetized it through scale. What’s different this time is that capital intensity is higher, and the payoff depends on whether future AI workloads actually monetize.
That said, we’re not bullish with our eyes closed. For us to change our view, we’d need to see at least one of a few things: the cloud companies revising CapEx guidance downward, orders being cancelled or pushed out, or AI product revenue and usage coming in below expectations. As of mid-2026, none of those has happened.
The ROI risk is real. But the facts, for now, sit on the bullish side. When the data starts to roll over, there will be time to change the call. We’re just not there yet.
Concern #2: Is this another 2000?
How did the 2000 bubble actually break? Demand back then was rising too. Internet users and traffic were growing year after year. What broke wasn’t the demand side. It was supply.
There was a popular line at the time: internet traffic was doubling every 100 days. Telecom companies bought that curve and started burying fiber along railways and highways. Fiber had a peculiar economic feature — once you’ve dug the trench, throwing in more cable barely adds cost. So they just stuffed in more than a decade of future capacity at once. Dozens of companies were digging their own trenches in parallel. Supply ended up vastly outrunning demand. The fiber that got laid down crashed to the floor in price, and by the time real traffic caught up to fill it, it was a decade later. The companies that put it in the ground didn’t survive to see that day.
This cycle has its share of froth too. Big cycles always do. Some companies are just slapping “AI” on themselves, and some of today’s spending will look reckless in hindsight.
But on the supply side, this round is the opposite of 2000. AI doesn’t just need a trench. Transformers are custom heavy equipment, gated by silicon steel supply and very long permitting timelines. Grid interconnection can’t be parallelized the way trenching could — you wait in line behind the public grid, often for several years. And most importantly, electricity can’t be pre-laid the way fiber was. You can’t bury ten years of future power capacity in the ground today and wait for the demand to show up.
So the way 2000 broke is very hard to replicate here.
The AI Show Isn’t Over. It’s Just Intermission.
Over the past few days, SpaceX has pulled back sharply from its highs and even broken below its IPO-day close. The market is nervous again. Watching this many giants line up at the capital window, it’s easy to start wondering whether AI has peaked.
We don’t read it that way.
The reason the giants are raising money in size right now is that the story has more acts ahead — and the obstacles get harder, not easier, from here. Look at the five clouds: not a single 2026 CapEx guide has been revised down. They’ve all been revised up. Look further out: transformers take four years to deliver, and connecting a data center to the grid means waiting in line for several years more. These aren’t obstacles that get solved by just throwing more money at them.
So this wave of fundraising looks dramatic. But really, it’s intermission.
Don’t call the top yet. The AI show isn’t over. It’s just turned to a new act.
TIDE Capital
TIDE Capital is an actively managed hedge fund investing across global public markets. Our team brings together veterans from leading investment banks, hedge funds, and technology firms. We pair the analytical rigor of traditional capital markets with the execution edge of modern quantitative methods.
website: tidecap.com
mail: info@tidecap.com
twitter: twitter.com/tidecap
medium: tidecap.medium.com
Disclaimer
The information and data presented in this article are obtained from public sources, and TIDE Capital makes no guarantees regarding their accuracy and completeness. Any predictions, speculations, or opinions contained in this article are statements about future events and may differ significantly from actual results due to limitations in data timeliness, assumption validity, uncertainty factors, and unforeseeable risks. Any advice and opinions in this article are for reference purposes only and do not constitute recommendations to buy or sell any digital assets. They do not constitute investment advice or solicitations. The strategies that TIDE Capital may adopt may be the same, different, or unrelated to those inferred by readers based on this article. Investors should carefully consider any decisions and seek appropriate legal and financial advice when necessary. Any misunderstanding or misuse of the content in this article does not constitute the responsibility of the author or the publishing institution.
