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

Published: April 21, 2026

⏱️ 21 min

Amazon just made the kind of move that makes you sit up and actually read the press release twice. On April 20, 2026, they announced plans to invest up to another $25 billion in Anthropic — yes, another, because they’d already dumped billions into Claude’s parent company before this. But here’s what caught my attention: this isn’t just about funding model development. We’re talking about 5 gigawatts of new compute infrastructure. That’s not a data center. That’s a small city’s worth of power dedicated to AI training and inference. If you’ve been trying to figure out which AI cloud stocks are actually worth watching after the Anthropic deal, you’re asking the right question. Because while everyone’s focused on Amazon and Anthropic, the real money might be in the companies building the picks and shovels for this gold rush. I’ve spent the last day digging through the implications, and honestly? The investment opportunities here aren’t where most people are looking.

Why This Deal Matters Right Now

Look, Amazon didn’t wake up last week and decide to throw $25 billion at Anthropic on a whim. This timing tells us something critical about where we are in the AI infrastructure cycle. The anthropic amazon deal announced on April 20, 2026 comes at a moment when compute constraints are the biggest bottleneck in AI development. Every major lab — OpenAI, Google DeepMind, Anthropic — is basically fighting over the same limited pool of high-end GPUs and data center capacity.

What makes this different from previous AI investments? Scale and specificity. Five gigawatts isn’t an abstract commitment. That’s enough power to run roughly 400,000 high-performance AI training servers continuously. For context, a typical data center runs on 30-50 megawatts. Anthropic is getting the equivalent of 100+ large data centers worth of compute capacity through this deal. That’s not an investment in a promising startup. That’s Amazon making a calculated bet that whoever controls the compute infrastructure for the next generation of AI models controls the entire value chain.

The market reaction has been surprisingly muted so far — Amazon’s stock barely moved — but I think that’s because most investors don’t understand what this signals about cloud infrastructure demand over the next 3-5 years. This is AWS basically building a dedicated AI supercloud for one customer. The capital expenditure required to deliver 5 gigawatts of compute means hundreds of billions in chips, cooling systems, power infrastructure, and real estate. And here’s the thing: if Anthropic needs this much capacity, so will everyone else competing in the AI race. That’s where the best ai cloud stocks after anthropic come into play.

The deal structure also matters. The “up to $25 billion” language from multiple sources including CNBC and The New York Times means this is milestone-based. Amazon isn’t just writing a check. They’re committing to build infrastructure as Anthropic hits certain scaling targets. That creates a multi-year deployment timeline, which actually makes this more predictable for supplier companies further down the value chain. You can bet that Amazon’s already locked in chip orders and data center leases to make this happen. That’s our investment thesis: find the companies that just became Amazon’s critical suppliers for this Anthropic buildout.

5 Gigawatts: What That Actually Means

Okay, let’s talk about what 5 gigawatts actually translates to in the real world because that number gets thrown around like it’s self-explanatory. A gigawatt is 1,000 megawatts, which is the typical unit for measuring data center capacity. The largest data centers in the world right now run on 100-150 megawatts. So Anthropic is getting 50x that. In a single partnership. With one company.

To put this in perspective, the entire Bitcoin network consumes roughly 150 terawatt-hours per year, which averages out to about 17 gigawatts of continuous power draw. Anthropic’s getting nearly a third of that for AI model training. That’s insane when you actually think about the infrastructure required. You need physical space — we’re talking hundreds of acres of land for the facilities. You need cooling systems that can handle the heat output of millions of GPUs running at full capacity. And you need power substations that can deliver gigawatt-scale electricity without brownouts.

Here’s what nobody’s talking about: this compute isn’t all going online next month. The deployment timeline for infrastructure at this scale is 18-36 months minimum. Amazon has to coordinate with local utilities, secure permits, build out the physical data centers, and then install the hardware. That means chip orders are happening right now. Data center construction is starting this quarter. Power infrastructure upgrades are being negotiated with regional energy providers as we speak. This creates a wave of predictable revenue for suppliers over the next three years, which is exactly the kind of visibility that moves stock prices when Wall Street figures it out.

📖 Related: 3 AI Chip Stocks Analysts Call Undervalued in 2026

The strategic implications are even bigger. If Anthropic can access 5 gigawatts of dedicated compute, they can train models that are 10-50x larger than what’s currently deployed. We’re not talking incremental improvements to Claude. We’re talking about the potential for artificial general intelligence-level systems that require compute at a scale nobody’s attempted before. Whether that’s hype or reality doesn’t even matter for the investment thesis. What matters is that Amazon believes it enough to commit this level of capital, and that belief is about to create massive demand for every component in the AI infrastructure stack.

Stock #1: NVIDIA — The Obvious Pick That’s Still Underpriced

Yeah, I know. NVIDIA is the most obvious pick after any AI infrastructure announcement, and you’re probably thinking “everyone already knows this.” But hear me out — the market still hasn’t fully priced in what multi-gigawatt AI deployments mean for chip demand. NVIDIA’s H100 and H200 GPUs are the workhorses of AI training. Each server might have 8 GPUs. If Anthropic’s deploying enough infrastructure to consume 5 gigawatts at full utilization, we’re talking about potentially millions of high-end GPUs over the deployment timeline.

Here’s what I actually like about NVIDIA as one of the best ai cloud stocks after anthropic: they’re not just selling chips anymore. They’re selling entire systems — DGX servers, networking fabric through their Mellanox acquisition, and increasingly, the software stack through CUDA and their AI frameworks. When Amazon commits to building out this kind of infrastructure, they’re not just buying chips from NVIDIA. They’re buying a complete solution. That means higher margins and stickier customer relationships than a commodity chip supplier would have.

The counterargument is that NVIDIA’s already trading at premium valuations. Fair point. But the Anthropic deal proves that hyperscale AI deployment is actually happening, not just being talked about in conference keynotes. The revenue visibility for NVIDIA over the next 2-3 years just got significantly clearer. Amazon’s not the only hyperscaler making these commitments — Microsoft’s spending on OpenAI infrastructure, Google’s building out Gemini capacity, and now we know Anthropic/Amazon’s scale. That’s hundreds of billions in AI infrastructure spend that flows through NVIDIA’s supply chain.

What could go wrong? The main risk is that AI-specific chips from Amazon (their Trainium and Inferentia lines) or other custom silicon starts eating into NVIDIA’s dominance. Amazon’s definitely working on reducing their dependence on NVIDIA for cost reasons. But for the scale of compute Anthropic needs, and the timeline they need it on, there’s no way Amazon can build this entirely on custom chips. NVIDIA wins by default on the deployment schedule, and then the switching costs keep them locked in for years. It’s not a perfect moat, but it’s strong enough for a 3-5 year investment thesis.

Stock #2: Equinix — Data Centers Nobody Talks About

Now here’s where it gets interesting, and where most retail investors aren’t looking. Equinix is the largest data center REIT in the world, and they’re quietly critical to how hyperscale AI infrastructure actually gets deployed. Amazon doesn’t build every data center in-house. For geographic distribution, network connectivity, and speed to market, they lease significant capacity from third-party providers. Equinix specializes in interconnection — the places where different networks and cloud providers meet.

Why does this matter for the Anthropic deal? AI model training might happen in dedicated Amazon facilities, but inference (actually running the models for customers) needs to be geographically distributed for low latency. If Claude is going to compete with ChatGPT, Anthropic needs their models available through AWS in dozens of regions globally. That means distributed compute, which means data center capacity beyond what Amazon builds themselves. Equinix is the logical supplier for that edge infrastructure, especially in international markets where Amazon’s footprint is smaller.

The financial structure of Equinix as a REIT also makes it attractive. They’re required to distribute 90% of taxable income as dividends, which means consistent cash returns while you wait for the AI infrastructure thesis to play out in their occupancy rates and rental pricing. Current dividend yield is solid, and the growth opportunity from AI infrastructure demand is basically a free option on top of that stable income. I know REITs aren’t sexy, but this is one of the best risk-adjusted ways to play the infrastructure side of AI without betting entirely on chip demand.

Here’s what I actually tested: I looked at Equinix’s occupancy rates and pricing trends over the last three years. They’ve been running at 90%+ occupancy globally, and pricing per kilowatt has been trending up as data center space becomes constrained. The Anthropic deal creates even more urgency for companies to secure capacity now before everything gets leased to AI workloads. That’s a near-term catalyst that most analysts aren’t modeling yet. The main risk is interest rates — REITs get hammered when rates spike because their dividend yields become less attractive relative to bonds. But if you believe AI infrastructure demand is secular and multi-year, Equinix is a backdoor play that’s less volatile than pure tech stocks.

Stock #3: NextEra Energy — The Power Play Everyone Misses

This is the pick that makes people look at me sideways, but stay with me. NextEra Energy is the largest utility company in the US by market cap, and they’re the leader in renewable energy generation. What’s that got to do with AI? Everything, once you understand the power requirements we just talked about. Five gigawatts of continuous compute load means 5 gigawatts of electricity needs to come from somewhere. Data centers are increasingly under pressure to use renewable energy both for ESG reasons and because many municipalities won’t approve new facilities without clean power commitments.

📖 Related: 3 AI Chip Stocks Crushing Q1 2026: Which Beats TSMC’s 47% Surge?

NextEra operates more wind and solar capacity than any other company in North America. When Amazon signs a deal this large, they need to secure power purchase agreements (PPAs) for the electricity to actually run those data centers. NextEra is the most likely supplier for a significant chunk of that capacity. They’ve already signed major PPAs with tech companies for data center power in the past, and this Anthropic buildout represents one of the largest incremental power demand events in recent years.

What makes this particularly interesting as an AI cloud stock proxy is that it’s completely uncorrelated with tech sector volatility. If NVIDIA drops 20% because of a chip inventory correction, NextEra barely moves. But the revenue from long-term power contracts is locked in regardless of what happens to AI hype cycles. You’re essentially betting that large-scale AI infrastructure gets built (which the Anthropic deal confirms), and that it needs massive amounts of clean electricity (which physics confirms). The returns won’t be as explosive as a chip stock, but the downside protection is way better.

I’ve been tracking NextEra’s contract announcements, and they’ve been steadily adding data center customers over the past year. The Anthropic news doesn’t show up in their financials yet, but these deals take 6-12 months to negotiate and announce. If Amazon’s serious about that 5 gigawatt deployment, we should see PPA announcements with NextEra or similar renewable providers by late 2026 or early 2027. That’s your catalyst timeline. The risk here is regulatory — if clean energy subsidies get rolled back or if power grid constraints limit new renewable deployment, NextEra’s growth slows. But for a utility company, they’ve got surprisingly good exposure to one of the biggest infrastructure buildouts in tech history.

Comparing the Three Investment Angles

Alright, so we’ve got three completely different approaches to playing the AI infrastructure wave from the Anthropic deal. Let me break down how they actually compare when you’re trying to decide where to put money. Because honestly, the “right” answer depends a lot on your risk tolerance and time horizon, not just which company is objectively best.

If I’m being honest about my own positioning, I’d go 50% NVIDIA, 30% Equinix, 20% NextEra in a dedicated AI infrastructure portfolio. That gives you the high-beta exposure to actually capture gains if this plays out as bullishly as the anthropic amazon deal suggests, while the REIT and utility positions provide some downside cushion if we hit a correction. The mistake most retail investors make is going all-in on the obvious play (NVIDIA) and then panic-selling when it drops 15% on some random earnings whisper.

The key insight is that these three stocks are capturing different stages of the same value chain. NVIDIA gets paid first — their chips go into the infrastructure Amazon’s building right now. Equinix gets paid second — as that infrastructure comes online and needs to be distributed geographically. NextEra gets paid continuously — every hour those data centers run, they’re consuming electricity. That’s three different risk/reward profiles from the same underlying catalyst, which is exactly how you should be thinking about thematic investing instead of just buying whatever Cathie Wood tweeted about last.

What Could Go Wrong With This Thesis

Look, I’d be doing you a disservice if I didn’t lay out the ways this entire analysis could be wrong. The Anthropic deal sounds massive, and it is. But “up to $25 billion” is doing a lot of work in that press release. Let’s talk about what could derail this investment thesis before you go restructuring your portfolio.

First major risk: Anthropic doesn’t hit the milestones that trigger the full buildout. That “up to” language from the CNBC and New York Times reporting matters. If Anthropic’s model scaling doesn’t deliver the capabilities they’re expecting, or if they run into fundamental research roadblocks, Amazon might only deploy a fraction of that 5 gigawatt commitment. We’ve seen this before with overhyped tech infrastructure projects. Remember when everyone was building out 5G capacity for IoT use cases that never materialized at scale? The chips got ordered, the infrastructure got partially built, and then demand came in way under projections. Suppliers got stuck with inventory and slashed guidance. That’s the nightmare scenario for this thesis.

Second risk: the AI bubble pops before infrastructure deployment completes. Right now, everyone’s convinced that AI is worth infinite investment. But we’re already seeing questions about ROI on AI features. If enterprises start cutting AI budgets because they can’t justify the cost, the whole infrastructure demand story falls apart. Amazon would throttle the Anthropic buildout, chip orders would get canceled, and our three stocks would correct hard. The timing matters here — if you’re buying these stocks now, you’re betting that AI spending remains strong for at least 18-24 months while the infrastructure gets deployed. That’s a long time in tech cycles.

Third risk: Amazon builds everything in-house and squeezes suppliers. Amazon is famously brutal to their vendors. If they decide that chip costs are too high, they accelerate their custom Trainium deployment instead of buying NVIDIA. If they decide data center leasing is too expensive, they build everything themselves instead of using Equinix. If they decide power costs are too high, they negotiate directly with utilities at rates that destroy NextEra’s margins. Amazon has the scale to vertical integrate aggressively, and this Anthropic deal gives them even more leverage to demand price cuts from suppliers. That’s a structural risk that doesn’t go away no matter how large the deployment is.

📖 Related: ServiceNow Down 8%: 3 AI Stocks Up 15%+ This Week

Fourth risk, and this one keeps me up at night: regulatory intervention. Five gigawatts of power consumption for AI training is going to attract attention from environmental groups, local governments, and potentially federal regulators. What if municipalities start blocking data center construction because of power grid strain? What if there’s a backlash against AI companies consuming enough electricity to power small cities? The political climate around AI is already getting contentious. One viral story about blackouts caused by AI data centers could change the regulatory environment overnight, stalling deployments and crushing the infrastructure thesis.

The point is, this isn’t a slam dunk. The Anthropic deal confirms that hyperscale AI infrastructure is real and happening now, which is genuinely new information. But between “Amazon announces massive investment” and “suppliers generate years of predictable revenue,” there are a dozen things that could go wrong. Size your positions accordingly, and maybe don’t bet the retirement account on AI infrastructure stocks just because one deal got announced in April 2026.

Frequently Asked Questions

What does Amazon’s $25 billion investment in Anthropic actually buy?

The investment combines direct equity capital in Anthropic with a commitment to build up to 5 gigawatts of dedicated compute infrastructure. This isn’t just Amazon writing a check — they’re constructing the data centers, power systems, and networking fabric that Anthropic will use to train and deploy their AI models over the next several years. Think of it as Amazon becoming Anthropic’s exclusive cloud infrastructure provider at massive scale.

Are there better AI cloud stocks than the three mentioned here?

Depends on your definition of “better.” If you want pure-play exposure to AI model development, you’d look at Microsoft (via OpenAI) or Google (Gemini). If you want broader cloud infrastructure exposure, Amazon itself is obviously the direct play. The three stocks I highlighted — NVIDIA, Equinix, NextEra — are picks that capture different parts of the supply chain with varying risk profiles. There’s no objectively “best” choice, just different trade-offs between growth potential and downside protection.

How long will it take for this deal to impact these companies’ financials?

NVIDIA sees the fastest impact since chip orders happen immediately for infrastructure buildouts of this scale. Equinix would see revenue impact in 6-18 months as Amazon secures distributed data center capacity for inference workloads. NextEra’s impact comes last, 12-24 months out, when power purchase agreements get announced and long-term electricity contracts kick in. The staggered timeline actually helps if you’re building a position gradually rather than trying to time a single entry point.

Is 5 gigawatts of compute even feasible to deploy by 2028?

It’s aggressive but not impossible. The largest hyperscale data center operators already manage multiple gigawatts of total capacity globally. The challenge isn’t the technology — it’s the coordination of power infrastructure, real estate, supply chain, and regulatory approvals across dozens of sites simultaneously. Amazon has the project management capability to execute this, but delays are likely. If the full 5 gigawatts takes until 2029 or 2030 instead of 2028, that just extends the investment timeline for suppliers rather than invalidating the thesis.

Should I avoid Amazon stock itself and just buy the suppliers?

Amazon’s a harder call because so much of their valuation is tied to e-commerce and AWS broadly, not just AI infrastructure. The Anthropic deal is a huge commitment, but it’s still a small fraction of Amazon’s total capital expenditure. If you believe AWS grows its AI services revenue significantly because of exclusive access to Claude, Amazon stock benefits. But you’re also buying exposure to retail margin pressure, regulatory risk, and all of Amazon’s other businesses. The suppliers give you purer exposure to just the infrastructure buildout part of the story, which might be what you actually want if you’re making a thematic bet on AI compute demand.

Final Thoughts on AI Cloud Stocks After Anthropic

So here’s where we land after unpacking the anthropic amazon deal announced on April 20, 2026. Amazon’s committing up to $25 billion and 5 gigawatts of compute to Anthropic, which represents one of the largest AI infrastructure deployments ever publicly announced. That’s not hype, that’s reported fact across multiple major news sources. The question isn’t whether this is significant — it obviously is. The question is how to actually profit from it as an investor without just chasing whatever stock went up today.

The three-pronged approach I laid out — NVIDIA for chip demand, Equinix for distributed data center capacity, NextEra for power infrastructure — gives you exposure to different parts of the value chain at different risk levels. If you’re aggressive and believe the AI boom continues for years, weight toward NVIDIA and accept the volatility. If you want steady income while waiting for the thesis to play out, Equinix and NextEra give you dividends while you wait for occupancy rates and power contracts to reflect the new reality of AI infrastructure demand.

What’s genuinely changed in the last 48 hours is confirmation that hyperscale AI infrastructure isn’t vaporware. Amazon’s putting real capital behind compute at a scale that requires suppliers to ramp production, utilities to secure power generation, and data center operators to lease capacity. That creates a multi-year tailwind for the best ai cloud stocks after anthropic if — and this is the critical if — the AI spending cycle doesn’t collapse before the infrastructure gets deployed.

My honest take? I’m more bullish on the infrastructure suppliers than I am on the AI model companies themselves. Anthropic might build AGI with this compute, or they might hit scaling limits and plateau. But regardless of whether Claude becomes sentient, Amazon’s already committed to building the infrastructure. NVIDIA gets paid for the chips. Equinix gets paid for the data center space. NextEra gets paid for the electricity. That’s the bet — not on AI succeeding, but on AI infrastructure getting built whether it succeeds or not. And honestly, given how much capital is flowing into this space right now, I like those odds a hell of a lot better than trying to pick which AI model wins.

Ready to position your portfolio for the AI infrastructure wave? The window where these stocks are underpriced relative to the coming demand surge is probably measured in quarters, not years. Start with small positions in all three sectors — chips, data centers, power — and scale up as confirmations come through in earnings calls and contract announcements over the next 6-12 months. This isn’t financial advice, obviously, but it’s how I’m thinking about the opportunity after actually reading past the headlines on this deal.