Tech Is Laying Off Workers and Crowning AI Kings. Here's Who Sells the Shovels.

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 dynamic is playing out right now in tech, and prediction markets are painting a remarkably clear picture of it.

Betting markets currently price an 87% chance that tech layoffs in 2026 will exceed 2025 levels. At the same time, the AI race is consolidating hard. Anthropic holds a 46.5% probability of being named the best AI model by year's end, while OpenAI sits at just 10.5% and xAI at 13.5%. SpaceX has an 87.5% chance of not going public before 2027. Tesla's Optimus robot has only a 25.5% chance of a commercial launch this year. Paramount acquiring Warner Brothers carries an 80.5% probability. And the Nasdaq falling below 19,000 by December? Only 16.5%.

Put these numbers together and a story emerges. The tech sector is simultaneously shrinking its workforce and funneling capital into fewer and fewer AI winners. The economist Ray Dalio has written about what he calls the productivity paradox, where transformative technology can destroy jobs even as it creates enormous value for a small number of companies. That is exactly what prediction markets are describing in 2026.

The Barbell: A Few Winners, Many Losers

The market implication is bearish for broad tech but bullish for AI infrastructure leaders. When 87% of bettors expect accelerating layoffs and media companies are merging to survive streaming economics, you're looking at a sector that is rationalizing. Cutting heads. Consolidating. The gains from AI are real, but they are not being spread around. The S&P 500 is priced at roughly a coin flip to stay above current levels despite the biggest technology investment cycle in decades. That tells you the market understands these AI gains are concentrated, not distributed.

This creates what you might call a barbell outcome. On one end, a narrow group of mega-cap AI leaders keeps outperforming. On the other end, the typical tech company faces margin compression from tariffs, rising energy costs, and a shrinking workforce that signals weakness, not strength. The middle gets squeezed.

The investment question becomes: who benefits no matter which AI company wins the crown?

Selling Shovels in the AI Gold Rush

The strongest plays in this pattern are the companies that supply infrastructure to every AI lab, regardless of whether Anthropic, OpenAI, or xAI comes out on top. Think of them as the modern equivalent of Levi Strauss selling jeans to miners.

CEG (Constellation Energy) is the highest-conviction infrastructure play. AI data centers consume staggering amounts of electricity, and Constellation is the largest nuclear power generator in the United States, providing the 24/7 carbon-free baseload power that hyperscalers specifically demand for their ESG commitments. Microsoft already signed a 20-year nuclear power purchase agreement with Constellation to restart a unit at Three Mile Island. Whether Anthropic or OpenAI wins the AI race, both need the lights on. Nuclear capacity is constrained, which gives Constellation pricing power. The infrastructure relevance score here is 82 out of 100, and the signal is a strong buy at 78% confidence.

NVDA is the canonical shovel-seller. Every frontier AI lab trains on NVIDIA GPUs. The company's datacenter segment is running above $47 billion annualized, and the CUDA software ecosystem, which is the programming framework developers use to run calculations on NVIDIA chips, functions as a deep competitive moat with over 80% market share in AI training. The concentration pattern is actually bullish for NVIDIA because fewer, better-funded winners means larger GPU cluster purchases. Infrastructure relevance score: 91 out of 100. Buy signal at 72% confidence.

ANET (Arista Networks) provides the high-speed networking switches that connect GPU clusters inside data centers. Every AI facility, whether built by Microsoft for OpenAI, Google for DeepMind, or Amazon for Anthropic, needs Arista's 400G/800G Ethernet switches. With three or more separate massive AI infrastructure buildouts happening simultaneously, Arista is arguably the purest networking shovel-seller available. They hold roughly 35% market share in cloud data center switching, ahead of Cisco in hyperscale deployments. Infrastructure relevance score: 85. Buy at 77% confidence.

VRT (Vertiv) handles something most people never think about: keeping data centers from overheating. As AI models scale up, the power density per server rack is increasing five to ten times over, making thermal management the critical bottleneck. Vertiv provides liquid cooling systems, power management, and IT infrastructure for data centers. They're one of only three or four major players in precision cooling. Infrastructure relevance score: 82. Buy at 78% confidence.

EATON is the electrical backbone, supplying power distribution units, transformers, switchgear, and UPS systems. Every data center needs reliable power distribution no matter which AI company occupies it. Eaton benefits from both the $100 billion-plus data center build cycle and the broader grid modernization required to support it. More diversified than Vertiv, which provides downside protection but limits pure AI upside. Infrastructure relevance score: 68. Buy at 76% confidence.

PWR (Quanta Services) goes one level deeper than the data center itself. They build the transmission lines, substations, and underground conduit that physically connect new AI facilities to the power grid. Training frontier AI models requires hundreds of megawatts, and the grid must be physically upgraded to deliver it. Quanta is the largest specialty electrical contractor in the U.S. Infrastructure relevance score: 62. Buy at 74% confidence.

EQIX (Equinix) operates as the neutral meeting point where AI labs, hyperscalers, and enterprises interconnect. As a REIT, a real estate investment trust that must distribute most of its income as dividends, it provides yield support during uncertain times. Long-term data center leases mean tech layoffs are essentially irrelevant to Equinix's revenue. Infrastructure relevance score: 78. Buy at 70% confidence.

VST (Vistra) complements Constellation as a diversified power generator with nuclear, natural gas, and battery storage assets. Its Texas footprint is strategically important given the concentration of data center construction in that state. Infrastructure relevance score: 75. Buy at 71% confidence.

The Platform Winners

MSFT wins regardless of which frontier model dominates because Azure hosts Anthropic, OpenAI through its partnership, and third-party models. Azure AI revenue is growing over 100% year-over-year. Even if OpenAI loses share to Anthropic, Microsoft benefits as the cloud substrate underneath. Buy at 68% confidence.

AMZN is the single largest infrastructure beneficiary if Anthropic wins the AI race at that 46.5% probability, given Amazon's $4 billion-plus investment and AWS hosting. AWS Bedrock offers model-agnostic AI infrastructure, making Amazon a true shovel-seller at the hyperscaler level. Amazon's logistics automation also creates internal AI productivity gains that partially hedge the employment contraction thesis. Buy at 65% confidence.

AVGO (Broadcom) supplies custom AI accelerators, like TPUs for Google, along with networking silicon and the VMware enterprise software stack. As capital concentrates into fewer AI winners, Broadcom supplies the silicon to nearly all of them. Buy at 75% confidence.

The M&A Play and the Hedge

WBD (Warner Bros. Discovery) is a direct merger arbitrage opportunity. With an 80.5% probability that Paramount acquires Warner Brothers, acquisition premiums typically offer 15-30% upside from pre-announcement levels. WBD trades at depressed valuations due to its heavy debt load and streaming losses, meaning the acquisition provides a floor value. Buy at 72% confidence.

SQQQ is the hedge leg of this barbell. It's a 3x inverse leveraged ETF on the Nasdaq 100, meaning it goes up roughly three percent for every one percent the Nasdaq falls. With 87% probability of accelerating tech layoffs and the median tech company facing margin compression, a small tactical position provides asymmetric protection. But this comes with a deliberate low confidence of 55% because leveraged instruments decay rapidly over time, and the market itself is pricing only 16.5% probability of a meaningful Nasdaq decline. This is a hedge, not a conviction bet.

Plays Considered and Rejected

SMCI (Super Micro Computer) designs AI server systems and has genuine relevance to the pattern, but its confidence is discounted to only 48% due to accounting irregularities, delayed SEC filings, and Ernst & Young's resignation as auditor in 2024. These are serious governance red flags that could impair the thesis regardless of how strong the fundamentals look.

AME (AMETEK) provides precision electronic instruments used in semiconductor manufacturing and data centers. It's a second-derivative play on the automation theme, with only an estimated 10-15% direct AI exposure. Weak buy at 65% confidence.

IRDM (Iridium) was considered as a satellite play tangentially connected to the SpaceX-stays-private theme but honestly assessed as too far removed from the core AI consolidation pattern. Infrastructure relevance score of just 28. Neutral, and not recommended as a core position.

The Self-Reinforcing Cycle

This is the part that should make the pattern click. What prediction markets are describing is a feedback loop:

1. AI investment concentrates into a few frontier labs (Anthropic, OpenAI, xAI)
2. These labs spend massively on compute infrastructure (GPUs, networking, power, cooling)
3. AI productivity tools eliminate the need for large tech workforces
4. Companies cut workers, improving margins for the few winners
5. Freed-up capital flows back into more AI investment
6. The cycle repeats, with infrastructure providers benefiting at every turn

This is Dalio's productivity paradox playing out in real time. The technology is transformative. It's also deflationary for employment. And the companies selling shovels to both sides of that equation are the ones positioned to win.

The Risks Are Real

Every play in this pattern carries meaningful risk, and ignoring them would be dishonest.

Valuation is the common thread. NVDA trades at roughly 35x forward earnings. VRT is above 40x after a 300%+ run in 2024. ANET sits around 35x. These prices assume sustained hypergrowth, and any stumble in AI capital spending could trigger sharp corrections. Historically, NVIDIA inventory correction cycles have caused 30-40% drawdowns.

China export restrictions could limit NVIDIA's addressable market. AMD's MI300X is gaining traction as a cost-competitive alternative. Custom silicon from Google, Amazon, and Meta could displace merchant GPUs over three to five years.

For WBD, antitrust regulators could block the Paramount deal. The combined entity would hold enormous content market share, and there's a 20% probability the deal simply fails, reverting WBD to its distressed standalone narrative. The company's $40 billion-plus debt load may also reduce any acquisition premium.

For power plays like CEG, nuclear plant operational risk is real. Unplanned outages impair revenue. Political risk around nuclear subsidies under the Inflation Reduction Act could shift with changing administrations. VST faces Texas grid volatility, where events like Winter Storm Uri represent genuine tail risks.

For the SQQQ hedge, daily rebalancing causes significant decay over holding periods longer than two weeks. The mega-cap AI winners that dominate QQQ weighting could mask broad weakness, making the instrument frustratingly unprofitable even if the thesis is correct. And Fed rate cuts could provide broad tech support that overwhelms the bearish signal.

Labor shortages in skilled electrical trades could constrain PWR. Hyperscalers building their own campuses could reduce reliance on EQIX. Rising interest rates compress REIT valuations. Customer concentration is a recurring vulnerability across AVGO, ANET, and SMCI.

Why This Matters for Your Money

If you have a 401(k) with a target-date fund or an S&P 500 index fund, you already own most of these companies. The question is whether you're positioned for a world where AI gains flow to a narrow set of winners while the broader tech workforce shrinks.

The prediction market data suggests that 2026 is not a year where a rising tide lifts all boats. It's a year where a few boats have rocket engines and everyone else is taking on water. If your retirement portfolio is heavily weighted toward broad tech exposure, understanding this concentration dynamic matters. The companies building the physical infrastructure of AI, the power generators, the chip makers, the cooling systems, the networking gear, are the toll roads of this transition. Traffic pays the toll regardless of where it's headed.

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