Tech Is Firing People and Hiring Robots: How to Invest in the AI Arms Race Without Picking a Winner

Something strange is happening in the tech industry right now. Prediction markets show an 87.1% probability that tech layoffs will accelerate this year. At the same time, the race to build the best AI is more intense than ever. Companies aren't shrinking. They're reshaping. They're cutting human headcount to pour more money into artificial intelligence, and the betting markets are giving us a surprisingly clear map of where that money is flowing.

This is a classic creative destruction cycle. The term comes from economics and it describes what happens when an industry tears down its old structures to build new ones. Think of it like a homeowner gutting a kitchen to renovate. The house looks worse in the middle of the project, but the end result is worth more. Tech companies are gutting their non-AI teams to fund the AI arms race, and the numbers tell a compelling story.

The AI Leaderboard According to Betting Markets

Prediction markets currently give Anthropic (the company behind Claude) a 50% chance of having the best AI model by December 2026. Google comes in at 28%, and OpenAI, the company that started this whole wave with ChatGPT, has fallen to just 10%. That's a remarkable shift. OpenAI's probability actually dropped 3.5 percentage points in a single 24-hour period.

Meanwhile, a few other tech storylines are playing out in the background. Tesla's Optimus robot has only an 18% chance of going on sale by the end of the year. The SpaceX IPO keeps getting pushed further out, with just a 4% chance of happening by May and 25% by June. And the Musk vs. Altman lawsuit? Prediction markets put the probability of a Musk-favorable outcome at about 34%.

All of this paints a picture of a tech sector that is simultaneously contracting in employment and concentrating its bets on AI. The major stock indices might look stable because mega-cap companies pouring money into AI offset the pain of broader layoffs. But underneath, there's enormous dispersion, meaning the gap between winners and losers within tech is widening fast.

The Gold Rush Logic: Sell Shovels, Not Gold

During the California Gold Rush, most prospectors went broke. The people who got rich were the ones selling pickaxes, shovels, and denim jeans. The same logic applies here. You don't need to know whether Anthropic, Google, or OpenAI wins the AI race. You just need to own the companies that supply all of them.

This infrastructure thesis runs deep, from the chips that power AI training all the way down to the electricity that keeps data centers running. Here's how the money flows:

1. AI labs like Anthropic, Google, and OpenAI need massive computing power to train and run their models.
2. That computing power comes from GPUs (graphics processing units repurposed for AI math) and custom chips.
3. Those chips are manufactured using specialized equipment in cutting-edge fabrication plants.
4. The chips get assembled into server racks and connected by high-speed networking gear.
5. All of it sits in data centers that need cooling systems, power distribution equipment, and enormous amounts of electricity.
6. Companies fund all of this by cutting their non-AI headcount, which is exactly what the 87.1% layoff acceleration probability reflects.

Every step in that chain has a company (or two) that dominates it. That's where the investment opportunities live.

The Chip Makers and Their Tool Suppliers

NVDA (NVIDIA) is the most direct way to play the AI arms race. They supply the GPUs that every major AI lab uses for training and inference. Whether Anthropic's 50% probability holds up or Google's 28% grows, every token trained runs through NVIDIA silicon. The company holds over 80% market share in AI training chips, and their CUDA software ecosystem creates a moat that competitors struggle to cross. Confidence here is 78%, the highest of any primary play. The risk? Valuation is stretched at roughly 30x forward revenue, and customers like Google and Meta are increasingly designing their own chips. If AI spending turns out to be a bubble, NVIDIA faces the steepest fall.

AVGO (Broadcom) designs custom AI accelerators for Google's TPU networking, Meta, and other hyperscalers, plus they dominate the networking silicon that connects GPU clusters together. Their AI-related revenue is now 35-40% of the total and climbing. Confidence is 80% with an infrastructure relevance score of 85 out of 100. Risks include customer concentration in Google and Meta, plus the ongoing integration of their VMware acquisition.

Further upstream, ASML holds what might be the most remarkable monopoly in all of technology. They are the only company on Earth that makes EUV lithography machines, the tools needed to print the tiny circuits on advanced AI chips. TSMC, Samsung, and Intel all need ASML's equipment. There is literally no alternative supplier. Infrastructure relevance score: 88 out of 100. The risks are geopolitical, including China export restrictions that limit their addressable market, and the cyclical nature of semiconductor equipment orders.

AMAT (Applied Materials) and LRCX (Lam Research) round out the fab equipment layer. Applied Materials handles deposition, etching, and inspection. Lam Research is especially interesting because they dominate equipment for HBM (High Bandwidth Memory), the memory stacks that sit directly on every NVIDIA GPU and represent a key bottleneck in AI chip production. Both carry BUY signals with confidence levels of 74% and 71% respectively.

The Cloud and Networking Layer

GOOGL (Google) is a dual play. They're both an AI competitor at 28% probability and an infrastructure provider through Google Cloud. They have DeepMind, proprietary TPU chips, and the distribution advantage of Search, Android, and Cloud. The creative destruction cycle actually helps Google because they can redirect resources from legacy projects to AI while keeping ad revenue flowing. Confidence: 72%. The main risk is antitrust action and the possibility that AI-powered search cannibalizes their own advertising model.

AMZN (Amazon) benefits through AWS, which is the primary cloud partner for Anthropic thanks to Amazon's $4 billion+ investment. AWS collects what amounts to infrastructure rent. AI training and inference workloads are scaling exponentially, and AWS capacity utilization benefits directly. The 87% layoff acceleration probability actually helps Amazon's cost structure as talent costs normalize. Confidence: 72%. Risks include aggressive competition from Azure and Google Cloud compressing margins, and the possibility that massive AI capex commitments become liabilities if spending slows.

MSFT (Microsoft) gets a WEAK BUY at 62% confidence, and the reasoning is important. Microsoft bet over $13 billion on OpenAI, which prediction markets now rate at just 10% for having the best AI. If Anthropic truly displaces OpenAI, Microsoft's investment faces serious impairment risk. Azure still has value as infrastructure, but the narrative that drove Microsoft's 2023-2024 AI gains may be unwinding. Copilot monetization remains unproven at scale.

ANET (Arista Networks) makes the high-speed switches that form the connective tissue of AI clusters. AI training requires massive data traffic between GPUs, and Arista's 400G and 800G switches are essential. Confidence: 76%.

The Physical Layer: Power and Cooling

This is where the infrastructure thesis gets really interesting because it's the layer most investors overlook entirely.

VRT (Vertiv) provides cooling systems, power distribution, and backup power for data centers. Every AI training cluster generates enormous heat and needs reliable electricity. Vertiv is one of three dominant players in this space, and their entire business is data center infrastructure. Confidence: 77%, infrastructure relevance score: 82.

VST (Vistra) and CEG (Constellation Energy) represent the electricity generation layer. Vistra operates natural gas, nuclear, and solar plants that supply power directly to hyperscaler data centers. Constellation is the largest nuclear power operator in the US and has signed landmark power agreements with Microsoft, including a deal to restart part of Three Mile Island. Nuclear is uniquely suited for AI data centers because it provides 24/7 carbon-free baseload power, exactly what companies need when their AI clusters run around the clock. Confidence levels: 70% and 69% respectively.

EATON gets a WEAK BUY at 70% confidence. They make switchgear, transformers, and UPS systems essential for data center buildouts, but their business is more diversified across industrial and commercial markets, which dilutes the AI exposure.

EQIX (Equinix), the world's largest colocation data center operator, also gets a WEAK BUY at 63% confidence. They're the real estate layer of the AI stack, but their REIT structure makes them sensitive to interest rates, and hyperscalers increasingly build their own campuses.

A Note on SMCI

One ticker worth mentioning is Super Micro Computer (listed under WDFC in our data but referring to SMCI). They assemble AI server racks, which makes them a logical infrastructure play. But serious accounting irregularities, auditor changes, and delayed SEC filings create risks that are too high to recommend with real money. The business model is correct, but governance risk makes it uninvestable for now. If their audit issues resolve cleanly, this becomes a buy. Until then, it's a NEUTRAL at just 45% confidence.

The Risks You Need to Understand

The biggest risk across all of these positions is the same one: what if AI investment turns out to be a bubble? If companies are spending hundreds of billions on AI infrastructure and the returns don't materialize fast enough, every name on this list gets hit. NVIDIA would face the worst multiple compression. The equipment makers would see orders evaporate. The power companies would be left with excess capacity.

Beyond the bubble risk, several specific threats deserve attention:

• Valuation: Nearly every name here is already priced for significant AI upside. You're not buying hidden gems. You're buying into a known trend at known-to-be-expensive prices.
• Geopolitics: China export controls limit the addressable market for NVIDIA, ASML, AMAT, and LRCX. Taiwan risk is existential for the semiconductor supply chain.
• Customer concentration: Many infrastructure companies depend on a handful of hyperscalers. When your top four customers represent the majority of your revenue, a single lost contract changes everything.
• Competition from within: Google, Amazon, and Meta are all designing custom AI chips. Every chip they build in-house is one they don't buy from NVIDIA.
• Interest rates: Capital-intensive infrastructure companies, from Equinix to Constellation Energy, are sensitive to borrowing costs.

Why This Matters for Your Money

If you have a 401(k) or any exposure to broad stock market index funds, you already own these companies. The question is whether the current allocation reflects what's actually happening. When prediction markets say there's an 87% chance tech layoffs accelerate while simultaneously showing an intensifying AI race, that tells you something important about where corporate spending is going.

The creative destruction happening in tech right now is the kind of structural shift that reshapes portfolios over five to ten years. Companies that supply the infrastructure for AI, the chips, the networking gear, the power plants, the cooling systems, stand to collect tolls on every dollar spent in this arms race. They're the modern equivalent of selling shovels during the Gold Rush.

The catch is that the market already knows this. Valuations are stretched across the board. The opportunity isn't in discovering something nobody knows. It's in understanding the structure of the bet: owning the infrastructure layer means you don't need to guess which AI company wins. You just need the race to keep going.

Analysis based on prediction market data as of April 2, 2026. This is not investment advice.