 |
| By Jurica Dujmovic |
Last week, I dropped a revelation: The AI bubble is already shrinking.
But I also warned you not to expect a dot-com style pop and crash.
Instead, a funding drought will likely lead to a quiet deflation.
And once the air has been let out, only a few AI firms will be left standing.
To be clear, funding will still go into AI. But who gets the money is becoming a smaller group as we speak.
If you didn’t read last week’s update, I suggest you start there.
Because this week, I’m diving into why we’ve begun to see this venture bifurcation … and who will survive it.
The Three-Tier Market: Who Survives, Who Dies and Why
The AI market isn't a monolith heading for uniform collapse.
It's stratifying into three distinct tiers, each facing radically different fates.
Understanding which tier a company occupies tells you everything you need to know about its survival odds.
Tier 1: The Hyperscalers
His tier includes your tech giants like Microsoft, Google, Amazon, Meta and Apple. These companies are essentially unassailable.
And here’s the kicker: None are making significant AI profits yet. Instead, they've found other ways to pay for their AI expansions.
For example, some are cross-subsidizing AI experiments with proven revenue streams as tech sector profit margins average 26% in 2024.
- Microsoft's Azure runs at an $75 billion annual rate.
- Though Meta's Reality Labs lost $11.5 billion in nine months of 2023, its Family of Apps business generated $41.8 billion in profit — fully funding the AI experiments with billions to spare.
In short, these hyperscalers can weather extended periods of disappointing AI returns because their core businesses print money.
And for the necessary capital not covered by other revenue streams, some firms have chosen to use debt and private credit for infrastructure.
Oracle, for example, aims to raise $18 billion in debt to build out its cloud infrastructure.
Meta and Morgan Stanley pitted asset managers against one another for the right to invest roughly $29 billion to build an AI data center in Louisiana.
And Global Infrastructure Partners (GIP) — a partnership between BlackRock, Microsoft and MGX — announced that NVIDIA and xAI will join the AI Infrastructure Partnership (AIP) to invest in new and expanded AI infrastructure.
For giants like these, the real risk here isn't failure. It's overbuilding.
This is where the dotcom parallel can become instructive: Telecom companies invested over $500 billion from 1996-2001, laying tens of millions of miles of fiber optic cable.
Even four years later, 98% remained unused.
Those companies went bankrupt. But that infrastructure enabled Netflix, cloud computing and the mobile internet.
And the value got captured entirely by these different players.
Today's data center buildout will most likely follow the same pattern: The builders lose, the applications win.
Tier 2: The Unicorns in Limbo
This is where the drama unfolds.
This middle tier contains popular AI firms that have massive war chests and genuine technical capabilities. This includes names like OpenAI, Anthropic and Scale AI.
But at the same time, these firms face existential questions.
Can they achieve returns justifying stratospheric valuations?
Can they compete against both Tier 1 hyperscalers with infinite resources … and the substantially cheaper, yet comparatively powerful models found in Tier 3 (more about that in a moment)?
Can they retain key talent when Meta offers $5-$10 million packages?
Over the next 18-24 months, these questions will be answered for each Tier 2 company.
That’s when we’ll begin to see the survivors rise to the top. And they’ll do that by demonstrating one of three things:
- a path to profitability at scale,
- a defensible moat preventing replication, or
- strategic value making them acquisition targets. Without one of those, their valuations will compress violently.
Tier 3: Everyone Else
This is where the vast majority of AI startups are. And it’s here where we can already see the casualties of the funding drought.
The shift from "innovation budgets" to "core IT budgets" means enterprises now demand ROI rather than experimentation.
And MIT's research suggests the application-layer startups that don’t have proven use cases will fail to achieve product-market fit before capital runs out.
For this tier, the 2026-2028 window is critical.
Companies that raised funding in 2021-2022 at inflated valuations are now hitting their 3- to 5-year marks. And that means they’ll either need to be profitable now … or risk relying on further fundraising rounds.
But with venture capitalists demanding proof of returns, most won't secure funding.
Seed-stage deal volume is down 43% year-over-year.
Down rounds — when a company raises new funding at a lower valuation than its previous financing rounds — are at decade highs.
And 95% of pilots fail to generate measurable returns.
In such an environment, it’s clear that these companies face a harsh uphill battle.
Simultaneously, enterprises reaching the 2- to 3-year mark on AI initiatives will face internal pressure to demonstrate returns on tens of millions in spending.
This could lead to project cancellations and vendor consolidation.
And physical infrastructure constraints are expected to emerge in 2027, which will accelerate the winnowing.
Gartner warns that 40% of existing AI data centers will be operationally constrained by power availability by 2027.
Companies with locked-in capacity agreements will maintain growth trajectories.
But the startups that are still seeking cloud capacity or specialized hardware by then? They’ll likely face availability crunches and cost pressures.
What This Means for Your TradFi AI Plays
If you're invested in the Tier 1 companies, you can relax.
Your risk is confined to temporarily disappointing returns and maybe a 10%-20% correction if AI progress stalls.
But overall, these industry titans aren't going to zero. They're profitable fortresses that can fund decade-long bets.
If you're evaluating Tier 2 unicorns, you’ll want to scrutinize:
- revenue trajectory,
- customer concentration, and
- talent retention
To help you find the ones more likely to survive.
For example, OpenAI has lost key researchers to competitors who can offer more competitive salaries. That’s a red flag.
On the other hand, Anthropic has secured multi-year enterprise contracts. That’s a green flag.
And be aware that the valuations for these companies are based on heroic assumptions.
Don’t take them at face value.
Instead, you’ll want to be sure that your investment has the fundamental support to make those outcomes at least plausible.
If you're evaluating or backing Tier 3 startups, take note: The companies that will receive funding are the ones showing actual returns.
The due diligence checklist for this tier is brutally simple:
- Narrow focus
- Measurable ROI
- Existing revenue base
If a startup misses two of those three, your money is at risk.
AI Is Here to Stay … But Not Every AI Firm
Again, this needs to be said: This is an equity correction, not an existential crisis.
It will hurt: Billions in capital will evaporate. Thousands of startups will shut down. Tens of thousands of jobs will vanish.
But the financial system won't collapse.
The profitable companies will keep profiting. The infrastructure being built — and overbuilt — will eventually enable applications we haven't yet imagined.
In short, don’t expect an explosion. Look for the leak.
It’s already started: slowly, selectively and asymmetrically.
And by the time the media notices, the correction will be over. The survivors will consolidate. The infrastructure will remain.
And we'll move on to debating the next bubble.
The biggest question for you right now is whether your AI investment has what it takes to survive, or if its already sprung its own leak.
I hope you use this update as a starting guide for how to vet your AI investments to stay on the profitable side of this critical moment.
Best,
Jurica Dujmovic
P.S. AI helped my colleague and cycles expert, Juan Villaverde, boost his Crypto Timing Model.
This is the very same model that has accurately called the highs and lows of the past three bull markets.
And now, testing proves it has the power to generate AVERAGE gains of 6,630% for each crypto asset it tracks.
To learn how you can give your own portfolio this AI boost, don’t miss your LAST CHANCE to watch Juan’s latest crypto briefing here.
