When The AI Bubble Bursts, These Companies Will Hold Up
Investors are jittery. AI valuations have gone vertical, data center spending is in overdrive, and every earnings call now teases even larger GPU clusters. If this ten trillion dollar boom deflates, almost everything in tech will feel the pressure.
None of these giants are immune. But they are not structured equally.
The difference between a sharp correction and a catastrophic collapse lies in a company’s revenue architecture: Is their business based on one-time purchase orders, or on mission-critical, recurring subscriptions? That split creates two distinct risk groups — companies with big upside but heavy exposure if budgets tighten, and another set of companies built to cushion the downturn while still riding the broader AI expansion. Separately, see Sell Nvidia, Buy Intel Stock For Big Upside?
Image by kp yamu Jayanath from Pixabay
- How NVIDIA Stock Gained 60%
- Nvidia Stock: Is the AI Boom Hiding A Hardware Time Bomb?
- Dot-Com vs AI Bubble: Is It Different This Time?
- Better Value & Growth: NVDA, MU Lead Analog Devices Stock
- Catalysts That Could Propel NVIDIA Stock to the Moon
- Nvidia Built An AI Moat. Can Rivals Find The Drawbridge?
Google: The Utility
Alphabet (NASDAQ:GOOG) business runs on recurring human behavior. People search, click, and advertise in both booms and busts. “Gyms near me” and “best running shoes” don’t disappear because AI spending cools off
- At the same time, it can deploy AI improvements immediately across Search, Ads, YouTube, Maps, and Workspace. Better ranking, better ad relevance, and more efficient compute via TPUs all compound at Google scale.
- Google has millions of advertisers, extremely high switching costs, and a diversified revenue base.
- Moreover, it is one of Nvidia’s largest customers. If Google shifts more workloads to its own TPUs – which are apparently getting a lot better – Google saves money while Nvidia loses revenue.
- Its lower earnings multiple reflects its utility-like stability rather than speculative AI upside. In a downturn, that stability becomes an advantage.
Microsoft: Workflow Subscription
Microsoft (NASDAQ:MSFT) sells the fabric of day-to-day work: Outlook, Excel, Teams, Windows, Azure.
- These are mission-critical tools that companies cannot cut, even during recessions. Email doesn’t get turned off. Spreadsheets don’t get cancelled. This recurring subscription base gives Microsoft a resilient revenue floor.
- There’s tons of built-in AI optionality built in here as well. Copilot tools in Office, Azure’s AI services, GitHub’s developer AI, and the broader integration of LLMs into enterprise workflows expand ARPU and drive higher cloud consumption.
Apple: Consumer Ecosystem
Apple is the only Mag7 name that isn’t spending tens of billions on data centers and GPUs. Its business success is tied to consumer loyalty and device cycles, not B2B compute spending.
-
Apple owns a vast consumer ecosystem with powerful loyalty and pricing power. Its $1,000 phones, $500 watches, and $10 a month services, independent of what happens with the AI cycle.
- While the downside risk is limited, it could still be one of the biggest beneficiaries of AI breakthroughs and improvement, because of its 2 billion plus user base, a $100 billion plus digital services business, and positioning to provide on-device intelligence.
Oracle: Legacy Lock-In
Oracle is unglamorous but deeply entrenched. Once a bank, insurer, or government agency adopts Oracle databases, they tend to stay for decades.
- Replacing it is risky, expensive, and often operationally impossible. That legacy lock-in creates a stable revenue floor.
- While newer AI companies depend on landing new customers to justify their valuations, Oracle benefits from customers who cannot leave.
- Oracle has bet big on AI (capex $35 billion in FY’26), yet most of that spend is already pre-sold via massive take-or-pay cloud/GPU contracts with stiff penalties. These deals lock in high-visibility revenue even if broader AI capex stalls, while its entrenched legacy base keeps paying regardless.
The Exposed Models: Who Gets Hit First When AI Spending Pauses
These companies are tied closely to capital cycles, aggressive valuations, or volatile competitive dynamics. They benefit the most when the boom is roaring and feel the pain first when it fades. Their core weakness is the dependence on large, non-recurring capital orders from a handful of customers.
Capital Equipment
Frontier model training is a winner-take-all arms race today, but it might be a fundamentally upfront capital purchase. Once hyperscalers hit “enough compute” for the current generation of models, orders can dip.
- For example Nvidia – with 60% revenue growth, 70 gross margins and 50% net margins is likely to be vulnerable as competition mounts or spending slows. What’s more, Nvidia’s revenue is concentrated. Unlike say Google, Microsoft, or Apple, who have billions of customers, just two customers accounted for almost 40% of Nvidia sales in a recent quarter.
- Nvidia depends on one-time sales of extremely expensive chips. Its latest chips sell for $30,000 or more. If hyperscalers slow their buying—because they have “enough GPUs”—Nvidia feels it immediately. Demand is tied to capex cycles, which are volatile. A pause in data center build outs directly translates to revenue pressure and valuation re-rating.
- Similar trends with other AI hardware players like AMD, Marvell, and Super Micro Computer.
Valuation-Dependent Growth
-
Hot software players like Palantir build strong AI-enabled software, but the valuation is dependent on fast expansion of customers and large deployments. If IT budgets tighten or deals are delayed, the stock can re-rate sharply because it has little margin for disappointment.
-
Unlike Oracle, it doesn’t enjoy a decades-long lock-in.
The Safety Checklist: How to Identify the Most Durable Tech Companies
Three simple questions reveal resilience:
1. Does it serve millions of customers instead of relying on a few giants?
Google does. Nvidia does not.
2. Does it sell a subscription rather than a one-time box?
Microsoft does. Hardware assemblers do not.
3. Would customers break their business if they stopped using it?
Oracle fits this description. Most AI infrastructure vendors do not.
If the AI bubble bursts, most tech assets will get marked down. But companies with diversified demand, recurring revenue, and deep lock-in have real shock absorbers. They don’t escape the storm, but they face far less damage than the firms selling the plumbing beneath the boom.
The Trefis High Quality (HQ) Portfolio, with a collection of 30 stocks, has a track record of comfortably outperforming its benchmark that includes all three – the S&P 500, S&P mid-cap, and Russell 2000 indices. Why is that? As a group, HQ Portfolio stocks provided better returns with less risk versus the benchmark index; less of a roller-coaster ride, as evident in HQ Portfolio performance metrics.