June 27, 2026 · 12 min read
After a blistering 2023-2025 AI rally that sent mega-cap tech stocks to all-time highs, a meaningful pullback has investors asking the big question: is the AI trade over, or is this the best entry point in years? We break down NVIDIA, Meta, and Amazon after the drawdown — examining fundamentals, valuations, and what history says about tech corrections.
The AI trade has been the defining investment theme since ChatGPT launched in November 2022. Between January 2023 and mid-2025, the three stocks at the center of the AI narrative delivered staggering returns: NVIDIA gained roughly 900%, Meta tripled, and Amazon nearly doubled. Trillions of dollars in market capitalization were created in under three years.
The thesis was straightforward: AI was going to transform every industry, and the companies building the infrastructure (NVIDIA), monetizing it through advertising and engagement (Meta), and powering it through cloud computing (Amazon) would capture disproportionate value. And for the most part, they delivered — revenue growth accelerated, margins expanded, and earnings estimates were revised upward quarter after quarter.
But 2026 has introduced something the market hadn't experienced during the AI rally: genuine uncertainty. A combination of factors has driven the pullback:
The result: NVIDIA is down roughly 26% from its highs, Meta has pulled back about 16%, and Amazon has declined around 12%. The Nasdaq Composite itself has corrected roughly 10% from its peak. The question every investor is asking is whether this is a healthy correction within a secular trend — or the beginning of the end for the AI trade.
NVIDIA remains the most direct play on AI infrastructure. The company's data center revenue is running at an annualized rate above $180 billion, driven by overwhelming demand for its Blackwell GPU architecture. Even after a 26% pullback from highs, the stock is still up over 500% since the AI trade began in earnest.
NVIDIA's Blackwell architecture represents a generational leap in AI compute efficiency, delivering roughly 4x the inference performance of the prior Hopper generation at the same power envelope. Hyperscalers are in the middle of a massive Blackwell upgrade cycle, and NVIDIA's backlog extends well into 2027. The company has also begun shipping its next-generation Rubin architecture to select partners for testing, ensuring another upgrade cycle on the horizon.
The correction in NVIDIA was driven primarily by three factors. First, margin compression fears as Blackwell's more complex manufacturing raised costs. Second, China export restrictions reduced the total addressable market by an estimated $12-15 billion annually. Third, concerns about competition from custom AI ASICs designed by Google (TPUs), Amazon (Trainium), and Microsoft (Maia) — though none have meaningfully dented NVIDIA's market share yet.
NVIDIA's data center segment now accounts for over 85% of total revenue, up from around 60% just two years ago. The company has effectively transitioned from a gaming GPU company to an AI infrastructure company. Its CUDA software ecosystem, which has over 5 million developers, creates a moat that competitors find extremely difficult to replicate — even when they offer competitive hardware.
The company's networking business (acquired via Mellanox) is also growing rapidly, as AI data centers require specialized high-bandwidth interconnects to move data between thousands of GPUs. NVIDIA's InfiniBand and Spectrum-X networking products are seeing demand growth that rivals its GPU business.
At ~32x forward earnings, NVIDIA is trading at its cheapest valuation multiple since the AI trade began. For a company still growing revenue at 80%+ year-over-year with 73% gross margins, the risk/reward setup is arguably better now than it has been at any point in the past 18 months. See how NVIDIA compares to its closest competitor: NVDA vs AVGO comparison.
Meta's transformation from a social media company into an AI-first advertising and engagement platform has been one of the most impressive corporate pivots in history. The stock has pulled back roughly 16% from its all-time high near $740, but the underlying business is performing exceptionally well.
Meta's AI investments are paying off in its core advertising business in ways that are directly measurable. The company's Advantage+ AI-driven ad targeting system has improved advertiser return-on-ad-spend (ROAS) by an estimated 30-40% compared to two years ago. Reels monetization has reached near parity with Feed on a per-impression basis — a milestone that many analysts didn't expect until 2027. And the company's new AI-generated ad creative tools are reducing the cost of campaign creation for small businesses, expanding the total advertiser base.
The primary overhang on Meta's stock remains Reality Labs, which continues to burn approximately $18 billion per year with limited near-term revenue contribution. However, the market has largely priced this in — Meta's core advertising business, stripped of Reality Labs losses, would trade at roughly 18-19x forward earnings, making it one of the cheapest high-growth tech stocks on the market.
Meta's decision to open-source its Llama family of large language models has been strategically brilliant. By giving away its models for free, Meta has built an enormous developer ecosystem around Llama — over 350 million downloads to date — which provides valuable feedback, reduces the company's own model development costs through community contributions, and ensures that the AI ecosystem doesn't become solely dependent on closed-source providers like OpenAI or Anthropic.
More importantly, Llama's success has given Meta leverage in the AI infrastructure market. Cloud providers like AWS and Azure now offer Llama models as first-class options alongside proprietary alternatives, which strengthens Meta's AI brand and creates a flywheel of improvement and adoption. The costs of developing Llama are effectively subsidized by Meta's core advertising business, making it a nearly free option for Meta to maintain its AI relevance.
Meta's AI strategy is unique among the mega-caps because it's primarily defensive and monetization-focused rather than infrastructure-focused. The company is using AI to make its existing 3.3 billion daily active user base more valuable, rather than trying to build a new business from scratch. That makes it arguably the lowest-risk AI play among the three stocks discussed here. Compare Meta's setup directly to Amazon: META vs AMZN comparison.
Amazon is perhaps the most multi-faceted AI story among the mega-caps. The company is simultaneously the largest cloud provider (AWS), one of the biggest AI infrastructure investors (Trainium custom chips, Bedrock AI platform), and the world's largest e-commerce company using AI to optimize everything from logistics to advertising.
After a growth slowdown in 2023-2024 driven by cloud optimization cycles, AWS is re-accelerating. The segment is growing at approximately 19% year-over-year, driven primarily by AI workloads. Amazon Bedrock — the company's managed AI model service — is one of the fastest-growing products in AWS history, with thousands of enterprise customers using it to deploy foundation models from Anthropic, Meta, Mistral, and Amazon's own Nova models.
Amazon's custom silicon strategy is also maturing. Trainium2, the company's second-generation AI training chip, offers roughly 4x the performance of its predecessor and is being used internally to train Amazon's own models at a fraction of the cost of using third-party GPUs. While Trainium hasn't displaced NVIDIA GPUs for most external customers, it gives Amazon a structural cost advantage for its own AI workloads.
Amazon's advertising business is often overlooked in AI discussions, but it's arguably the company's most AI-dependent growth engine. At roughly $60 billion in annualized revenue and growing over 25% year-over-year, Amazon's ad business is now larger than YouTube's. AI powers everything from product recommendations to sponsored listing optimization, and the segment carries margins estimated at 50%+ — making it a significant contributor to Amazon's overall profitability improvement.
Amazon's profitability has improved dramatically over the past two years. Operating income has more than doubled as the company rationalized its logistics network, improved fulfillment efficiency through AI-driven automation, and benefited from the high-margin growth of AWS and advertising. The company's overall operating margin has expanded from roughly 5% in 2023 to approximately 11% in 2026 — and analysts expect further expansion as AI-driven efficiencies continue to scale across the business.
The combination of revenue growth re-acceleration and margin expansion creates a powerful earnings growth trajectory. Consensus estimates suggest Amazon's earnings per share could grow at 25-30% annually over the next three years — one of the fastest rates among mega-cap stocks.
At ~28x forward earnings, Amazon trades at a reasonable premium given the combination of AWS re-acceleration, advertising growth, and the company's massive AI infrastructure moat. The biggest risk is that the $78 billion in 2026 capex — the largest capital expenditure commitment of any company in history — may take longer to generate returns than investors expect.
How do these three AI leaders stack up against each other across the metrics that matter most? Here's the full comparison.
| Metric | NVDA | META | AMZN |
|---|---|---|---|
| Current Price (approx.) | ~$145 | ~$620 | ~$215 |
| 52-Week High | ~$195 | ~$740 | ~$245 |
| Drawdown from High | ~26% | ~16% | ~12% |
| Forward P/E | ~32x | ~24x | ~28x |
| Revenue Growth (YoY) | ~80% | ~22% | ~12% |
| Operating Margin | ~62% | ~38% | ~11% |
| Free Cash Flow Yield | ~2.1% | ~3.4% | ~2.8% |
| AI Revenue Exposure | ~85%+ | ~40% | ~35% |
| 2026 AI Capex (est.) | N/A | ~$42B | ~$78B |
| Analyst Consensus | Strong Buy | Buy | Strong Buy |
| Avg. Price Target | ~$175 | ~$710 | ~$245 |
Each stock offers a different risk/reward profile. NVIDIA has the highest AI revenue exposure and growth rate but also the steepest drawdown. Meta offers the best valuation on a forward P/E basis with the most mature AI monetization. Amazon provides the broadest diversification with AWS, advertising, and e-commerce, but also carries the heaviest capex burden.
A few key takeaways from the table: NVIDIA offers the most upside to its analyst price target (~21% upside from current levels), while Meta and Amazon offer roughly 15% each. NVIDIA also has the widest drawdown from highs, which means it has the most ground to recover but also potentially the most risk if the pullback deepens. On a valuation basis, Meta is the cheapest at ~24x forward earnings, making it arguably the best risk-adjusted play for conservative investors looking for AI exposure.
It's also worth noting the free cash flow yield differences. Meta's ~3.4% FCF yield is the highest among the three, reflecting its capital-light advertising model versus the massive infrastructure spending required by NVIDIA's supply chain and Amazon's cloud buildout. For income-oriented investors or those who value cash generation, Meta stands out.
The most common bear argument is that AI is following the same trajectory as the dot-com bubble: revolutionary technology, irrational exuberance, and an inevitable crash. But the comparison breaks down on closer inspection.
In 2000, the Nasdaq was dominated by companies with no revenue, no business model, and sky-high valuations based purely on "eyeball counts." Pets.com, Webvan, and eToys had combined revenue of less than $500 million with a combined market cap exceeding $10 billion. The P/E ratio of the Nasdaq Composite exceeded 200x at the bubble's peak.
In 2026, the AI leaders are among the most profitable companies in history. NVIDIA generates over $75 billion in annual free cash flow. Meta's operating margins have expanded to 38%. Amazon's operating income has more than doubled in two years. The Nasdaq's forward P/E of roughly 28x is elevated but nowhere near bubble territory.
A more relevant comparison is the 2022 tech drawdown, when the Nasdaq fell over 33% amid rising interest rates and fears that tech growth was peaking. Meta fell 77% from its highs. Amazon dropped 56%. NVIDIA declined 66%. Every single one of those stocks went on to set new all-time highs within 18-24 months as earnings growth re-accelerated.
The key lesson from 2022: pullbacks in high-quality tech stocks with genuine earnings growth tend to be buying opportunities — as long as the fundamental thesis remains intact. The question for 2026 is whether the AI thesis is still intact. We believe it is, but with important nuances.
The single most important distinction between genuine tech revolutions and bubbles is whether earnings follow the hype. In the dot-com era, earnings never materialized — the hype was based on theoretical future revenue that never arrived. In the current AI cycle, earnings have not only materialized but consistently exceeded expectations. NVIDIA has beaten consensus EPS estimates for 10 consecutive quarters. Meta's earnings growth has re-accelerated after a post-COVID reset. Amazon's profitability is at all-time highs.
This doesn't mean AI stocks can't go lower — they absolutely can if macro conditions deteriorate or if AI spending slows. But it does mean that the fundamental underpinning of the AI trade is real revenue and real profits, not speculative hope. That's an important distinction when evaluating whether to buy, hold, or sell during a pullback.
Despite the pullback, several powerful structural drivers suggest the AI trade has years of growth ahead — not months.
According to McKinsey's 2026 AI survey, only about 28% of Fortune 500 companies have deployed AI in production at scale. The remaining 72% are either in pilot stages or haven't started. Enterprise AI spending is expected to grow from approximately $120 billion in 2026 to over $350 billion by 2030, implying a 30%+ compound annual growth rate. The enterprise adoption S-curve is still in its early innings.
To put this in perspective, cloud computing took roughly 15 years to move from early adoption to mainstream enterprise deployment. We are approximately 3 years into the AI deployment cycle. If AI follows even a modestly similar adoption curve, the biggest spending and revenue growth years are still ahead of us — not behind us. The companies that supply the infrastructure, platforms, and applications for this adoption wave (NVIDIA, Amazon, Meta) are positioned to benefit for years.
The first wave of AI spending was dominated by model training — building the largest foundation models required massive GPU clusters. But the second wave is about inference — running those models in production to serve billions of queries. Inference workloads are growing at an estimated 3-4x the rate of training workloads and require sustained, ongoing compute spending rather than one-time cluster buildouts. This is a recurring revenue tailwind for NVIDIA, AWS, and cloud providers.
Governments worldwide are investing in domestic AI infrastructure to reduce dependence on US cloud providers. The EU has committed over $20 billion to sovereign AI initiatives. Japan, India, Saudi Arabia, and the UAE are collectively investing another $40+ billion. This creates entirely new demand for NVIDIA GPUs and cloud infrastructure that didn't exist two years ago — and it's largely insensitive to private-sector capex cycles.
The shift from chatbots to AI agents — autonomous software that can browse the web, write code, manage workflows, and execute complex tasks — is creating a new category of AI revenue. Companies like Salesforce, ServiceNow, and Microsoft are embedding AI agents into enterprise workflows, driving incremental AI infrastructure demand. This "agentic AI" wave is still in its earliest stages.
Model capabilities continue to improve at a pace that surprises even AI researchers. The gap between AI systems and human-level performance on complex reasoning, coding, and analytical tasks has narrowed dramatically. As AI becomes more capable, the total addressable market for AI applications expands — creating a virtuous cycle of investment, improvement, and adoption.
Each new capability frontier unlocks new use cases that weren't previously viable. Multimodal AI (combining text, image, video, and code understanding) is enabling applications in healthcare diagnostics, autonomous driving, drug discovery, and financial analysis that would have been science fiction just three years ago. This expanding capability frontier is the reason AI spending continues to accelerate rather than plateau.
While we believe the AI trade has legs, investors should monitor several risk factors that could signal a more prolonged downturn.
If the 10-year Treasury yield moves above 5%, the valuation framework for growth stocks fundamentally changes. Higher discount rates compress the present value of future cash flows, which disproportionately impacts high-P/E tech stocks. Watch for sustained upward moves in bond yields as a warning sign for AI stock valuations. The Federal Reserve's monetary policy path remains uncertain, and any indication that rate cuts are delayed beyond 2027 would put additional pressure on growth stock multiples.
Hyperscalers are spending a combined $250+ billion on AI infrastructure in 2026. If quarterly earnings calls start showing disappointment in AI revenue monetization relative to this spend, expect significant multiple compression. The key metric to watch is each company's "AI revenue per dollar of AI capex" — and whether that ratio is improving or declining quarter over quarter.
Any indication that hyperscalers are slowing their AI infrastructure buildout would be immediately negative for the entire AI supply chain, starting with NVIDIA. Watch for language changes in earnings call guidance from Microsoft, Google, Amazon, and Meta regarding their 2027 capex plans. A deceleration from the current $250B+ pace would be a significant red flag.
If open-source models continue to close the performance gap with proprietary models, the value of AI infrastructure spending may shift from model providers to end-user applications. This would be negative for NVIDIA's pricing power and hyperscaler spending, but potentially positive for companies that use AI to improve their core business (like Meta's advertising).
Further tightening of AI chip export controls to China, or new restrictions on AI technology transfer, could meaningfully impact revenue for NVIDIA and other semiconductor companies. Investors should monitor US-China relations closely, as any escalation in technology restrictions creates headline risk and could shave billions from NVIDIA's addressable market. Additionally, tariff uncertainty on broader tech imports adds another layer of macro risk to the sector.
After analyzing the fundamentals, valuations, and structural drivers, our view is clear: the AI trade is not over. The pullback in NVDA, META, and AMZN represents a re-pricing of expectations, not a fundamental deterioration in the AI thesis. Enterprise adoption is still early. Inference demand is accelerating. And these three companies are generating genuine, growing earnings — not dot-com vapor.
But here's the critical nuance: stock selection matters far more in 2026 than it did in 2023. During the initial AI euphoria, almost everything AI-adjacent went up. Going forward, the market will differentiate between companies that are actually monetizing AI and those that are merely spending on it. The three companies analyzed here — NVIDIA, Meta, and Amazon — are among the best positioned because they're both spending on AI and generating clear returns from it.
For investors with a 2-3 year time horizon, this pullback looks more like an opportunity than a warning. The AI infrastructure buildout is a multi-decade trend, and the companies at its center are generating the revenue and profits to justify their valuations — something that was decidedly not true during the dot-com era.
The most prudent approach for most investors is to avoid trying to time the exact bottom. Dollar-cost averaging into positions over a 3-6 month period allows you to take advantage of continued volatility while building meaningful exposure to the AI theme. If you're already holding these stocks, the fundamental picture does not support panic selling — every major data point suggests that AI revenue and adoption are still accelerating.
The AI trade isn't over. It's just getting harder. And harder is exactly when disciplined, research-driven investors have their edge.
Want to compare these stocks side by side with real-time data? Try our free comparison tools:
Disclaimer: This article is for informational purposes only and does not constitute financial advice. All investments carry risk, and past performance does not guarantee future results. Always do your own research and consider consulting a licensed financial advisor before making investment decisions. Data and prices are approximate as of the publication date.