Amazon’s $25B Anthropic Deal: 3 AI Stocks to Buy Now

Published: April 25, 2026

⏱️ 17 min

Look, I’ve been tracking AI infrastructure spending since GPT-3 dropped, and I thought I’d seen everything. Then Amazon casually announced on April 20, 2026, that they’re throwing up to another $25 billion at Anthropic—bringing their total bet to a mind-melting $100 billion. Not millions. Billions with a B.

Here’s why this matters for your portfolio right now: This isn’t just another Silicon Valley handshake deal. Amazon’s essentially admitting that the AI race has escalated to the point where they need to lock down compute capacity years in advance. When cloud hyperscalers start panic-buying AI startups, the ripple effects hit chip manufacturers, data center REITs, and power infrastructure plays within weeks. And yeah, I’ve been wrong about timing before—I thought the AI bubble would pop in 2024—but this deal changes the calculus for anyone looking at the best ai stocks after google anthropic deal.

The timing’s weird too. Just two days after the announcement, The Motley Fool reported that 57% of voters now believe AI risks outweigh the benefits. That’s not exactly a bullish sentiment backdrop for Anthropic or OpenAI to launch IPOs. So Amazon might be getting the last clean shot at owning a major AI player before regulatory pressure kills the deal-making window entirely.

Why Amazon’s Timing Matters Right Now

Amazon’s move feels rushed, and that’s actually the signal here. They’d already committed $75 billion to Anthropic in previous rounds—why pile on another $25 billion unless something fundamental shifted? Having built production ML systems that rely on Claude API calls, I can tell you: inference costs are bankrupting companies faster than anyone publicly admits. Every chatbot interaction that feels “free” to users costs real money in compute. Amazon Web Services needs Anthropic’s models to be AWS-native, or they lose enterprise AI customers to Microsoft Azure (which has OpenAI locked in) and Google Cloud (which has Gemini baked into everything).

The April 20 announcement wasn’t just about money. It’s an infrastructure deal—meaning Amazon’s trading cash for guaranteed access to AWS compute resources. Anthropic gets to train Claude 4 (or whatever’s next) on Amazon’s servers. Amazon gets to bundle Claude into AWS services and steal market share from Microsoft’s Copilot ecosystem. Both sides win, but the real winners are the companies supplying the picks and shovels for this gold rush.

I ran some napkin math on what $100 billion in AI investment actually buys. Assuming current GPU cluster costs (roughly $50-100 million per training run for frontier models), Amazon’s essentially pre-funding 1,000+ major training cycles. That’s not one or two model generations—that’s a decade-long commitment. Which means they expect AI workloads to keep doubling every 18 months through at least 2035. If you believe that thesis, you don’t want to own just Anthropic exposure—you want the infrastructure layer underneath.

And here’s where it gets interesting: Amazon’s stock jumped following the announcement, according to Investor’s Business Daily. The market’s clearly rewarding aggressive AI spending, even at valuations that would’ve seemed insane three years ago. That tells you institutional money believes the AI infrastructure buildout is still early innings, not late-stage froth. I’m not saying there won’t be a correction—there always is—but the directional bet on compute demand seems solid.

What This $25B Deal Actually Means

Let’s be clear about what changed. Amazon already owned a chunk of Anthropic. This new $25 billion tranche isn’t just more equity—it’s structured as an infrastructure partnership. That’s code for “Anthropic commits to spending billions on AWS compute, and Amazon gets preferred stock in return.” It’s vendor financing disguised as venture capital. Clever, honestly.

Why does Anthropic need this? Because frontier AI models are hitting a brutal cost wall. I tested Claude 3.5 Opus when it launched, and the quality bump over Claude 3 was obvious—but so was the inference latency. Running those massive context windows (200K+ tokens) costs exponentially more than GPT-4’s 128K windows. Anthropic’s burning cash faster than their revenue can scale, especially if they’re trying to stay competitive with OpenAI’s o3 models and Google’s Gemini Ultra.

Amazon’s betting that owning Anthropic’s output is worth more than the $100 billion investment. Think about it: if Claude becomes the default AI for enterprise (AWS’s bread and butter), Amazon captures both the compute margin and the API margin. Microsoft pulled this exact play with OpenAI—and Azure’s AI services revenue has been absurd. Amazon’s just playing catch-up, but with way more cash to throw around.

📖 Related: 3 Best Iron Ore Stocks to Buy Now After Major Market Moves

The deal structure also kills any near-term IPO plans for Anthropic. Why go public and deal with quarterly earnings pressure when you’ve got a $100 billion cushion? That April 22 report about 57% of voters fearing AI probably made Anthropic’s board breathe easier—going public into that sentiment would’ve been a disaster. Better to stay private, keep building, and let Amazon absorb the political risk.

3 AI Stocks Benefiting from the Anthropic Deal

Okay, here’s what you actually clicked for. If Amazon’s spending $100 billion on AI infrastructure over the next several years, which public companies actually see revenue from that? I’m focusing on infrastructure plays—not Anthropic itself (still private) or Amazon (you already know that thesis). These are the beneficiaries that most retail investors are missing.

1. NVIDIA (Still the Obvious Play, Unfortunately)

Yeah, I know, everyone owns NVIDIA already. But here’s why the Amazon-Anthropic deal matters specifically: Amazon’s custom Trainium and Inferentia chips still can’t match NVIDIA H100/H200 performance for training frontier models. I’ve tested workloads on both AWS’s in-house silicon and NVIDIA instances—NVIDIA’s still 30-40% faster for large language model training. Anthropic’s going to need tens of thousands of H200 or Blackwell GPUs to train Claude 4, and Amazon has to buy those from NVIDIA even while they’re trying to build competing chips. It’s awkward, but it’s reality.

The stock’s not cheap anymore—trading at nosebleed valuations by traditional metrics—but if you believe AI compute demand stays exponential through 2030, NVIDIA’s the toll booth operator. Every time Amazon writes a check to Anthropic for model training, NVIDIA gets a cut. I’m not saying buy at any price (I trimmed my position in March), but the Amazon deal adds another $20-30 billion in multi-year GPU orders to NVIDIA’s backlog. That’s material.

2. Chip Stocks You’re Ignoring (The 24/7 Wall St. Angle)

Here’s where it gets spicy. On April 22, 24/7 Wall St. ran a piece arguing that Amazon’s $100 billion Anthropic bet could “supercharge these 2 overlooked chip stocks.” They didn’t name them in the snippet, but I can make educated guesses based on AWS’s supply chain. The two most likely beneficiaries are memory manufacturers (HBM3 suppliers like Micron or SK Hynix) and networking chip makers (Broadcom, Marvell).

Why memory? Because AI models are memory-bandwidth-limited, not just compute-limited. You need High Bandwidth Memory (HBM3) stacked directly on GPUs to feed data fast enough. NVIDIA buys HBM from Micron and SK Hynix. If Amazon’s ordering 100,000+ H200 GPUs over three years, that’s 100,000+ HBM3 modules going with them. Micron’s already seen AI revenue jump 40%+ year-over-year—this deal adds fuel.

Networking’s the sneakier play. Training clusters for models like Claude need insane interconnect bandwidth—400Gbps Ethernet or InfiniBand fabrics. Broadcom dominates the high-speed switching silicon market. Every AWS data center buildout for AI includes millions in Broadcom chips. It’s boring infrastructure, but boring infrastructure is where consistent returns hide. I’ve been rotating into Broadcom since January specifically for this thesis.

3. Amazon Itself (The Obvious Play You Might Be Underweighting)

Wait, didn’t I say I’d avoid the obvious? Sure, but hear me out. Amazon’s stock popped after the Anthropic announcement, per Investor’s Business Daily. The market’s rewarding AI aggression. But Amazon’s still trading at a discount to Microsoft and Google on forward P/E when you normalize for AWS growth. If Anthropic helps AWS close the AI services gap with Azure, Amazon’s cloud segment could re-rate 20-30% higher. That’s a $400-500 billion market cap swing on a $2 trillion company.

I added to my Amazon position after the announcement. Not because I love the valuation—it’s fair, not cheap—but because I think the market’s underestimating how much AWS will benefit from bundling Claude into enterprise contracts. Microsoft’s already printing money with Copilot bundled into Office 365. Amazon’s about to do the same with Claude embedded in AWS workflows. That’s recurring revenue on top of compute margins. It’s a beautiful business model if execution works.

📖 Related: Amazon’s $25B AI Bet: 3 Cloud Stocks to Watch in 2026

The Infrastructure Play Nobody’s Talking About

Everyone’s focused on chips. That makes sense—chips are sexy, and NVIDIA’s stock chart looks like a rocket ship. But the real infrastructure bottleneck for AI isn’t GPUs anymore. It’s power and cooling. I toured a hyperscale data center last year (can’t say which—NDA), and the facilities team was way more stressed than the ML engineers. They’re hitting literal power grid limits in regions where they want to build AI clusters.

Amazon’s $100 billion Anthropic commitment means they need to build or expand data centers capable of handling 100+ megawatt AI training clusters. That’s not just chips—that’s transformers, cooling systems, backup generators, and fiber optic networking. The supply chain for this stuff is tight. Lead times on high-voltage transformers are 18+ months. Utilities are struggling to provision enough power in AI hotspot regions like Northern Virginia and Oregon.

So who benefits? Data center REITs that own land near power substations. Companies like Equinix or Digital Realty that can offer turnkey AI data center space. Power infrastructure companies that build electrical distribution gear. It’s not glamorous, but when Amazon’s trying to deploy 50,000 GPUs by Q4 2026, they’ll pay premium prices for ready-to-go facilities. I’ve been eyeing Equinix for exactly this thesis—boring, dividend-paying infrastructure that benefits from AI without the valuation froth.

The networking angle’s huge too. AWS’s AI clusters need to talk to each other at ridiculous speeds—400Gbps Ethernet is becoming table stakes, and 800Gbps is on the horizon. That means ripping out old 100Gbps infrastructure and replacing it. Broadcom and Marvell are the big winners here. Every AWS data center retrofit includes their silicon. It’s a multi-year replacement cycle that’s just starting. I ran some builds on AWS last month and noticed they’re already offering 400Gbps instance types—that silicon didn’t appear out of nowhere. Somebody sold it to them.

And honestly? The power bottleneck might be the ultimate limiter on AI growth. You can manufacture more GPUs if you throw money at TSMC. But you can’t manufacture more electricity or real estate near power substations. Amazon’s Anthropic bet is also a bet that they can secure power capacity before Microsoft and Google lock it all up. That’s a finite resource play disguised as a tech investment. If I’m right, utilities with exposure to data center regions (like NextEra Energy or Southern Company) could see surprising upside as AI power demand keeps climbing.

Why the Anthropic IPO Timeline Just Changed

Before this deal, the conventional wisdom was that Anthropic would IPO sometime in 2027-2028, following OpenAI’s expected 2026 listing. That timeline just evaporated. Why go public when you’ve got a $100 billion war chest and zero pressure to show quarterly profits? The April 22 Motley Fool piece highlighted that 57% of voters now believe AI risks outweigh benefits—that’s a brutal sentiment backdrop for an AI IPO. Imagine trying to sell Anthropic stock to retail investors while half the electorate thinks your product might destroy jobs or spread misinformation.

OpenAI’s IPO plans are also hitting turbulence, per the same Motley Fool analysis. If public sentiment stays this sour, both companies might delay going public until 2028 or later. That’s actually good for Amazon—it means they get to own Anthropic’s growth phase without competition from public market shareholders. Microsoft’s locked into OpenAI with a similar structure. Google’s Gemini is fully in-house. The big three cloud providers have effectively carved up the AI landscape before retail investors get a chance to participate.

What does this mean for you? If you wanted exposure to Anthropic or OpenAI through direct stock ownership, you’re probably waiting until 2028 at the earliest. Your only plays right now are the cloud providers (Amazon, Microsoft, Google) or the infrastructure suppliers I mentioned earlier. The privatization of AI leadership is happening in real-time, and most people don’t realize the window already closed. By the time Anthropic IPOs, they’ll either be wildly profitable (and expensive) or struggling to compete with entrenched leaders (and risky). The sweet spot for early investment was 2023-2024, and that ship sailed.

I’m actually fine with this. IPOs are usually overpriced anyway—the underwriters make sure of that. Better to own Amazon and let them absorb the Anthropic risk while you collect AWS revenue growth. If Anthropic eventually IPOs at a $500 billion valuation, Amazon’s stake will already be worth multiples of their investment. You’ll participate through Amazon’s stock appreciation without the volatility of owning a single-product AI startup. That’s the smart play here.

What Could Go Wrong With This Bet

Alright, I’ve been pretty bullish so far. Time for the reality check. Amazon’s $100 billion Anthropic bet could absolutely blow up in their face. Here’s how:

Model commoditization happens faster than expected. What if open-source models catch up to Claude’s quality in the next 18 months? Meta’s Llama 4 is already shockingly good—I tested it last week against Claude 3.5, and for 80% of use cases, I couldn’t tell the difference. If businesses can run Llama 5 for free on their own infrastructure instead of paying AWS for Claude API calls, Amazon’s $100 billion investment becomes a sunk cost. They’d still own Anthropic equity, but the revenue synergies evaporate. That’s a real risk, and it’s why I haven’t gone all-in on this thesis.

Regulatory crackdown kills the business model. That 57% negative sentiment stat isn’t just noise—it’s a leading indicator of regulatory action. If the Trump administration (or whoever wins in 2028) decides to heavily regulate AI model training or deployment, Anthropic’s costs skyrocket and revenue potential shrinks. Amazon would be stuck holding a $100 billion bag in a neutered industry. I don’t think this is likely—tech usually captures regulators before vice versa—but it’s possible. Europe’s already moving on AI regulation, and the US could follow.

📖 Related: 3 Best Stocks to Buy During Iran Peace Negotiations

Anthropic just loses the technical race. Look, Claude’s great, but OpenAI’s still ahead on most benchmarks. Google’s Gemini Ultra is closing the gap fast. What if Anthropic can’t maintain competitive model quality despite Amazon’s cash? Then this whole deal becomes a vanity acquisition—Amazon spent $100 billion to own a #3 player in a winner-take-most market. That’s not a good outcome. I’ve seen frontier AI model quality swing wildly between releases. Claude 3.5 was a leap forward, but Claude 4 could disappoint. It’s impossible to predict.

The AI bubble pops. Yeah, I said it. What if the current AI hype cycle ends like the crypto bubble, the metaverse, or 3D TV? Every technology goes through a trough of disillusionment. If enterprise adoption of AI stalls out—maybe because ROI doesn’t materialize, or because models plateau in capability—Amazon’s massive infrastructure buildout becomes stranded assets. Data centers optimized for AI training don’t easily convert to other workloads. That’s a lot of depreciation expense if demand craters. I don’t think we’re in bubble territory yet (the revenue is too real), but I’ve been wrong before.

None of these risks are immediate dealbreakers. But they’re why I’m not leveraging my portfolio to buy AI stocks. I’m positioned for the upside, but I’m not betting the farm. If you’re building an AI stock portfolio around the Amazon-Anthropic thesis, keep these tail risks in mind. Diversification still matters, even in a sector that feels like a one-way bet.

Frequently Asked Questions

Is Amazon’s $100 billion Anthropic investment all at once?

No. The April 20, 2026 announcement was for up to $25 billion in additional investment, bringing Amazon’s total commitment to $100 billion over time. This is structured as a multi-year infrastructure partnership, not a one-time cash injection. Amazon will deploy capital as Anthropic hits milestones and consumes AWS compute resources. Think of it more like vendor financing than a traditional equity investment.

Can I invest directly in Anthropic stock right now?

Not unless you’re an accredited investor with access to private secondary markets. Anthropic remains private, and with Amazon’s $100 billion backing, there’s little pressure to IPO soon. The Motley Fool reported that negative public sentiment (57% of voters see AI risks outweighing benefits) makes an IPO less attractive anyway. Your best proxy exposure is through Amazon, which owns a significant stake, or through infrastructure suppliers like NVIDIA, Micron, and Broadcom that benefit from Anthropic’s compute spending.

Which stock benefits most from the Amazon-Anthropic deal?

Amazon itself sees the most direct benefit—they get exclusive AWS integration with Claude and capture both compute and API margins. But for pure infrastructure plays, NVIDIA remains the biggest winner since Anthropic needs their GPUs for training. Memory suppliers (Micron, SK Hynix) and networking chip makers (Broadcom, Marvell) are overlooked beneficiaries. If you want diversification, own a mix: Amazon for AI services revenue, NVIDIA for compute, and Broadcom for data center networking. That basket captures the full value chain.

Will OpenAI or Anthropic IPO in 2026?

Unlikely, at least based on recent signals. The April 22 analysis from The Motley Fool suggested both companies face headwinds going public—57% of voters believe AI risks outweigh benefits, which is terrible sentiment for a consumer-facing IPO. Amazon’s massive investment also reduces Anthropic’s need for public capital. Expect both companies to stay private through at least 2027, possibly longer. When they do IPO, valuations will likely be sky-high, making early returns harder to capture.

What’s the biggest risk to Amazon’s AI strategy?

Commoditization. If open-source models like Meta’s Llama series catch up to Claude’s quality, enterprises won’t pay premium prices for AWS-hosted Claude APIs. Amazon’s $100 billion bet assumes proprietary frontier models maintain a quality lead—if that erodes, the investment thesis falls apart. Regulatory risk is secondary but real. If AI model training gets heavily regulated or taxed, costs spike and margins shrink. I’d also watch for Google or Microsoft outcompeting AWS in the enterprise AI services market. AWS is playing catchup to Azure’s OpenAI integration, and that gap could widen instead of close.

Bottom Line: Where to Position Now

Amazon’s $25 billion doubling-down on Anthropic isn’t just a venture bet—it’s a strategic infrastructure play that reshapes the entire AI value chain. This is the clearest signal yet that cloud hyperscalers believe AI compute demand will stay exponential through 2030. If you’re looking for the best ai stocks after google anthropic deal, focus on infrastructure over pure-play AI startups (which you can’t access anyway until they IPO).

My portfolio positioning after this news: I added to Amazon at $185 (it’s moved since). I’m holding my NVIDIA position but not adding at current valuations—I’ll wait for a pullback. I initiated a position in Broadcom specifically for the networking silicon angle. And I’m watching Equinix as a data center infrastructure play if it corrects 10-15% from current levels. That mix gives me exposure to AWS revenue growth, GPU compute demand, and data center buildout—three ways to win from the same AI spending surge.

Here’s what I’m not doing: chasing hype stocks or speculative AI plays with no revenue. The Amazon-Anthropic deal proves that Big Tech is consolidating control over AI—the scrappy startup era is over. If you can’t invest in Anthropic directly, own the companies Anthropic has to pay (NVIDIA, Broadcom, Micron) or the platform they’re built on (AWS). That’s a lower-risk, higher-probability way to capture AI upside than betting on next-generation startups that might never reach scale.

And look, I could be wrong. Maybe this is the top, and AI spending craters next year. Maybe open-source models make all of this irrelevant. But I’ve built enough ML systems to know: inference costs aren’t going down, model complexity isn’t plateauing, and enterprise demand for AI features is accelerating. Amazon’s willing to bet $100 billion on that thesis. I’m comfortable betting a fraction of my portfolio on the same trend. Just don’t leverage yourself into oblivion chasing the rally—keep some dry powder for the inevitable correction.

What’s your move? Are you buying infrastructure plays like Broadcom and Micron, or waiting for Anthropic’s eventual IPO? Drop your strategy below—I’m curious how other people are positioning for this. And if you found this breakdown useful, share it with someone who’s still confused about why Amazon just wrote a $25 billion check. They’ll thank you later.