The AI Economy Is Splitting in Two: Where Prediction Markets Say the Money Is Heading

Something unusual is happening in the tech economy, and prediction markets are making it visible in hard numbers. The industry is splitting into two halves. On one side, a tiny handful of AI leaders and platform monopolists are consolidating enormous power and spending record amounts on infrastructure. On the other side, the broader tech workforce is getting laid off at accelerating rates. This is not a temporary dislocation. It looks more like a permanent rearrangement of where value lives in the technology sector.

The numbers from betting markets paint a stark picture. Anthropic, the company behind Claude, holds a 42% probability of being named the best AI by end of 2026, and that number has been climbing, up 2.3% in just the last 24 hours. Google sits at 27%, xAI at 14%, and OpenAI, the company that kicked off the entire AI frenzy with ChatGPT, is surprisingly low at just 11%. Four players now account for roughly 94% of the probability in the "best AI" race. Meanwhile, tech layoffs have an 85.9% probability of exceeding 2025 levels, a SpaceX IPO carries an 88% probability before 2027 (with a 16% chance it happens before June 2026 and a 65% chance before July 2026), Elon Musk has a 75% probability of becoming a trillionaire by 2027, and Netflix is at 74% probability of raising subscription prices this year.

Read all of those signals together and a pattern emerges that the legendary investor Ray Dalio would call a "paradigm shift." Wealth and power are concentrating, not distributing. AI leaders are spending billions on chips and data centers while simultaneously eliminating human jobs. Platform monopolists like Netflix are raising prices because they can. And an enormous amount of capital is flowing toward a single individual's ecosystem of companies.

This creates a specific, investable thesis. If you remember the California Gold Rush, most of the miners went broke. The people who got rich were the ones selling shovels, pickaxes, and blue jeans. The same logic applies to the AI race. You don't need to correctly guess whether Anthropic or Google wins. You need to own the infrastructure that all of them must buy.

The Self-Reinforcing Loop

The pattern here works as a cycle that feeds itself:

1. AI leaders raise massive funding and generate revenue from their products.
2. They spend that money on GPUs, data centers, networking, and power infrastructure.
3. This spending accelerates AI capabilities, which displaces human workers.
4. Displaced workers and impressed customers drive more demand for AI products.
5. AI companies raise more funding and generate more revenue.
6. The cycle begins again, with each loop concentrating more value in fewer hands and more spending on infrastructure.

Companies in the middle, the mid-tier SaaS businesses, the enterprise software firms that aren't AI-native, are getting squeezed from both directions. They face disruption from AI products above and workforce costs that are harder to cut below. The prediction market data on layoffs suggests this squeeze is already well underway.

The Direct AI Plays

Most of the AI race leaders are private companies, which limits how you can invest directly. But two public companies stand out.

GOOGL is a BUY at 75% confidence. Google's 27% probability in the AI race reflects genuine competitiveness through DeepMind, Gemini, and its enormous cloud infrastructure. Unlike Anthropic (private), Google is the most directly investable AI leader. The important nuance is that Google doesn't even need to "win" the best-AI perception race. AI integration across Search, Cloud, YouTube, and Waymo creates monetization pathways regardless. DeepMind's research output remains world-class. The flat movement in prediction markets (0% change in 24 hours versus Anthropic's +2.3%) suggests some relative underperformance is being priced in, but the breadth of Google's business provides downside protection.

GOOG (the other share class) carries a slightly lower confidence WEAK BUY at 60%, reflecting the same thesis but recognizing that the DOJ antitrust case is a genuine overhang. If the government forces Google to divest Chrome, Android, or parts of Search, the business transforms fundamentally. Google's internal culture has also historically been slow to commercialize its research advantages, and the enormous capital expenditures required for AI are compressing near-term free cash flow.

TSLA is a WEAK BUY at only 55% confidence, and this one requires careful thinking. The 75% Musk trillionaire probability is driven largely by SpaceX and xAI, both private. Tesla is the only liquid stock in the Musk ecosystem, which means it acts as a proxy. The problem is that you're paying a massive premium for optionality on companies you can't actually own shares in. Tesla's auto fundamentals are deteriorating with declining market share, margin compression, and the distraction of Musk running SpaceX, xAI, Twitter, and his government work simultaneously. Robotaxi and Optimus timelines have been pushed back repeatedly. And there's an ironic risk: if SpaceX actually does IPO (remember, 88% probability), capital might rotate out of Tesla and into SpaceX directly.

The Shovels: Infrastructure That Wins Regardless

This is where the real thesis lives.

NVDA is a BUY at 80% confidence, the highest conviction call in the pattern. NVIDIA is the textbook shovel seller. Whether Anthropic, Google, xAI, or OpenAI wins the AI race is irrelevant because all of them must buy NVIDIA GPUs. The extreme concentration among AI leaders actually helps NVIDIA because fewer, richer winners spend more per company on compute. Data center revenue now accounts for roughly 88% of NVIDIA's total revenue, and the CUDA software ecosystem (the parallel computing platform that AI researchers use to write code for NVIDIA chips) creates a switching cost that competitors have struggled to replicate. The concentration in AI leadership means well-capitalized players are engaged in a capex arms race, which is precisely the demand pattern NVIDIA benefits most from. AMD's MI300X and custom chips from Google, Amazon, and Microsoft are genuine long-term threats, but the CUDA moat remains formidable today.

ASML is a BUY at 78% confidence, and this is the play that sits two levels upstream from AI. ASML makes the extreme ultraviolet lithography machines (essentially the most complex manufacturing tools ever built) that TSMC and Samsung use to produce the advanced chips powering all AI systems. There is literally no alternative supplier for cutting-edge EUV machines. Whether NVIDIA, AMD, or custom chips win the chip war, they all need chips made on ASML's machines. That is a monopoly position that is essentially impossible to replicate.

VRT (Vertiv) is a BUY at 76% confidence. Every AI data center needs power management, thermal management, and IT infrastructure, and Vertiv is one of the top three providers globally. Think of it as the electrical plumbing of the AI economy. AI models generate enormous heat, and cooling systems are becoming a critical bottleneck. Vertiv addresses that bottleneck directly.

ANET (Arista Networks) is a BUY at 75% confidence. As AI training clusters scale from thousands to hundreds of thousands of GPUs, the high-speed networking switches connecting them become a critical bottleneck. Arista provides that networking fabric. Meta, Microsoft, and the major cloud companies are all customers. The bifurcated tech economy actually helps here: fewer, larger AI data center operators means larger, more concentrated customers who buy at scale.

ETN (Eaton) is a BUY at 74% confidence. Eaton provides the transformers, switchgear, backup power systems, and power distribution equipment that every new data center requires. The tech layoff trend is particularly relevant here because companies are redirecting labor spending toward infrastructure spending, which flows directly to Eaton's order book.

CEG (Constellation Energy) is a BUY at 71% confidence and might be the most overlooked play in the entire pattern. AI data centers are consuming power at a scale that is straining the US grid, and nuclear is the only scalable, always-on, carbon-free power source that hyperscalers can contract for. Microsoft signed a 20-year agreement with Constellation to restart a reactor at Three Mile Island for this exact reason. Constellation is the largest nuclear power operator in the United States.

ORCL (Oracle) is a BUY at 69% confidence. Oracle Cloud Infrastructure has quietly become a preferred training cloud for AI labs that want an alternative to the AWS/Azure/Google Cloud oligopoly, reportedly including large xAI workloads. With xAI sitting at 14% in the AI race and growing, its infrastructure spend flows disproportionately to Oracle. Enterprise AI deployment also runs through Oracle's database layer, making it a two-way beneficiary.

AMZN is a BUY at 72% confidence. Amazon benefits through AWS, the cloud infrastructure layer that many AI companies, including Anthropic (in which Amazon has invested $8 billion), depend on for training and inference compute. Anthropic's surge to 42% probability of winning the AI race directly benefits Amazon as Anthropic's primary cloud partner. AWS acts as a shovel seller disguised as a primary player. Its margins expand as AI workloads increase regardless of which AI model ultimately wins.

EQIX (Equinix) is a WEAK BUY at 62-68% confidence. As the world's largest data center REIT (a real estate investment trust, meaning it owns the physical buildings and leases space) with 260+ facilities, Equinix provides the neutral ground where cloud providers, AI companies, and enterprises physically connect their networks. The REIT structure provides dividend income while you wait. The weaker conviction reflects the reality that hyperscalers increasingly build their own data centers rather than leasing from Equinix.

VST (Vistra) is a WEAK BUY at 60% confidence. Vistra is the largest competitive power generator in the US, with significant nuclear capacity after acquiring Energy Harbor and major natural gas generation. Its Texas exposure is particularly interesting: the Musk/SpaceX/xAI ecosystem is heavily concentrated in Texas, and the 75% Musk trillionaire probability combined with xAI's growing position suggests Texas-based AI infrastructure will scale significantly. Vistra is a direct power beneficiary of that geography. Lower conviction than Constellation due to commodity gas price exposure.

The Risks Are Real

Honesty about what could go wrong is essential.

For the AI leaders, antitrust action is a genuine threat. The DOJ case against Google could force structural changes to the company. Anthropic's widening lead raises the question of whether Google is losing the technical race despite outspending everyone. For Tesla, the risk list is long: declining auto margins, Musk distraction, brand damage from political involvement, robotaxi delays, and the possibility that a SpaceX IPO draws capital away.

For the infrastructure plays, the single biggest risk is that the AI capex cycle slows or reverses. If AI spending plateaus, NVIDIA faces a cyclical downturn. Vertiv's backlog could evaporate. Arista faces tough comps. ASML's semiconductor customers cut orders. Every infrastructure play in this pattern depends on continued, aggressive AI spending growth.

More specific risks include: NVIDIA's valuation at roughly 30x forward revenue leaves almost no room for disappointment. ASML faces geopolitical risk from China export restrictions. Constellation faces nuclear safety and regulatory uncertainty. Oracle's cloud infrastructure remains years behind AWS and Azure in capability. Eaton is a diversified conglomerate where AI exposure is diluted across aerospace and vehicle segments. Equinix faces rising interest rates that increase its cost of capital. And across the board, hyperscalers developing custom chips (Google's TPU, Amazon's Trainium, Microsoft's Maia) represent a long-term structural threat to NVIDIA's dominance.

The tech layoff data itself carries a second-order risk that most investors overlook. An 85.9% probability of layoffs exceeding 2025 levels means reduced consumer spending in tech-heavy cities like San Francisco, Seattle, and Austin. That has ripple effects on housing, local businesses, and tax revenue that don't show up in AI company earnings but affect the broader economy.

Why This Matters for Your Money

You don't have to be a tech investor for this pattern to affect you. If you have a 401(k) or any retirement account that holds an S&P 500 index fund, you already have significant exposure to the companies discussed here. The concentration of AI value in a handful of winners means your index fund performance is increasingly tied to whether these specific companies execute.

On the consumer side, the Netflix price increase probability (74%) is a small but telling signal. Platform monopolists are raising prices because they can. That dynamic extends beyond streaming to cloud services, software subscriptions, and eventually AI-powered tools that businesses pass costs for along to customers. The AI economy's value concentration doesn't just affect stock prices. It shows up in your subscription fees, your grocery delivery costs, and the prices you pay for services that are increasingly AI-powered behind the scenes.

The shovels-versus-gold framework matters for portfolio construction. If you're convinced AI will transform the economy but can't pick the winner, owning the infrastructure layer (chips, power, cooling, networking, data centers) gives you exposure to the trend without betting on a single horse. Every dollar Anthropic, Google, xAI, or OpenAI spends on their AI race flows through these infrastructure companies. That's the trade the prediction markets are pointing toward.

Analysis based on prediction market data as of March 20, 2026. This is not investment advice.