Article

Big Tech Is Burning $2 Billion a Day on AI — Here’s Why It Might Be Genius

Wall Street has a $700 billion question on its hands: Are the five biggest hyperscalers — Microsoft, Amazon, Alphabet, Meta, and Oracle — building the future, or torching shareholder capital at a rate of $2 billion per day?

On the surface, the number is staggering. $710 billion in AI-related capital expenditures this year alone, flowing into data centers, GPUs, fiber, energy infrastructure, and cooling systems. That’s more than the GDP of most countries. And every time an earnings report shows revenue growth that doesn’t perfectly match the capex pace, the “dot-com bubble 2.0” crowd comes out swinging.

But here’s what the skeptics keep getting wrong: they’re measuring today’s returns against tomorrow’s infrastructure. And when you actually run the math — carefully, with realistic assumptions — the spending starts looking less like recklessness and more like one of the greatest capital allocation stories in modern corporate history.

The AI revenue opportunity breaks down into three engines. First, consumer subscriptions — ChatGPT Plus, Claude Pro, Gemini Advanced — which could scale to roughly $120 billion annually as global penetration grows. That’s real money, but it’s the appetizer.

The main course is enterprise AI. There are 560 million knowledge workers globally representing $32 trillion in annual labor costs. AI doesn’t need to replace them — it just needs to automate 40% of their output. At a 20% value-capture rate (consistent with historical enterprise software pricing), that’s $2.56 trillion in annual revenue. And that’s the mid-case estimate.

Then comes physical AI — robots directed by the same intelligence models, deployed across manufacturing, logistics, agriculture, and healthcare. The global physical labor market runs $39 trillion annually. Robot-as-a-Service is already emerging (see Figure AI’s BMW deal, Amazon’s warehouse buildout, Tesla’s Optimus). Conservative projections put this at $4 to $5 trillion at maturity.

Add it all up: roughly $7 trillion in total addressable revenue. That’s 10x what the hyperscalers are spending on capex this year.

Now the critical question — margins. Consumer AI should converge toward 50-60% operating margins (it’s basically software). Enterprise AI sits around 35-45%. Physical AI runs lower at 20-30% because robots depreciate, need maintenance, and consume energy. Blended across the full revenue base, that’s roughly 32% operating margin, yielding about $1.77 trillion in annual after-tax profit at maturity.

For perspective: the entire S&P 500 currently generates about $1.5 trillion in annual after-tax earnings. The AI stack alone could surpass that, concentrated among maybe a dozen companies.

The calculated ROIC on cumulative net invested capital? Approximately 27-28% — right in line with Google and approaching Microsoft’s territory. That’s not a bubble. That’s generational compounding.

The catch? We’re deep in the J-curve right now. Returns won’t cross the 12% cost-of-capital hurdle until roughly year nine or ten. For the next several years, the spending will look ugly on a trailing-returns basis. And that’s exactly why the market is nervous — it’s pricing short-term pain into stocks that are playing a 20-year game.

But there’s a crucial distinction: the hyperscalers aren’t some overleveraged startups sweating their next funding round. Microsoft generates $90+ billion in free cash flow annually. Amazon has AWS. Alphabet has Search. Meta has advertising. They can fund this J-curve from existing operations without breaking a sweat.

The companies that should worry? The pure-play infrastructure borrowers financing AI buildouts with expensive short-duration capital. They’re the ones the J-curve will eat alive.

For investors, the takeaway is straightforward: the current choppiness in AI-related equities isn’t a sign that the thesis is broken. It’s what a J-curve looks like from the inside. And if the math is even directionally right, these prices will look like a gift in hindsight.

Article

Nvidia Crushed Earnings — Then Lost $260 Billion Anyway

Nvidia just delivered the kind of quarter most companies would frame and hang on the wall. Revenue of $68.1 billion, up 73% year-over-year. Net income of $43 billion — that’s $330 million in profit per day. Data center revenue alone hit $62.3 billion, accounting for 91% of total sales and blowing past the $60.7 billion Wall Street expected. First-quarter guidance came in at $78 billion, above even the most optimistic buyside whispers.

And the stock dropped 5.5%, erasing roughly $260 billion in market value in a single session. That’s more than the entire market cap of most S&P 500 companies — gone in a day.

Welcome to the new normal for Nvidia, where beating earnings is the bare minimum and anything short of a miracle gets punished. Over the past five quarters, the stock has tripled on the back of relentless beats. Options markets had priced in just 5.6% implied move — the lowest in three years — signaling that the “beat” was already baked in. Without a genuine upside surprise beyond the surprise everyone expected, there was nothing left to buy.

The real concern isn’t what Nvidia reported — it’s what it didn’t say. Investors wanted clarity on 2027 growth, and the earnings call didn’t deliver it. The $100 billion OpenAI partnership deal, once seen as a lock, now has “no assurance” language buried in Nvidia’s 10-K filing. Meanwhile, a broader anxiety is building around whether the AI capex boom is sustainable. Hyperscalers like Microsoft, Amazon, Google, and Meta are expected to spend over $1.1 trillion on AI infrastructure in 2026, but the downstream monetization hasn’t caught up. Bank of America’s latest fund manager survey flagged high AI capex as the second-largest systemic credit risk.

There’s also the inference question. As the industry shifts from training AI models (Nvidia’s sweet spot) to running them at scale, competitors could chip away at Nvidia’s dominance. Fundstrat’s Hardika Singh noted that Nvidia “missed on easing investors’ concerns about its narrowing moat in the evolving world of compute.” Jensen Huang pushed back, highlighting that the upcoming Vera Rubin architecture is specifically designed for inference workloads — but markets aren’t in the mood for patience.

Broadcom dropped 3%, Taiwan Semiconductor fell 2.8%, and the entire chip complex felt the gravity. The S&P 500 and Nasdaq both stumbled, weighed down by the sector that’s been carrying them for two years.

Here’s the thing, though: 61 out of 66 analysts still rate Nvidia a buy, with an average price target implying 37% upside. Janus Henderson’s Richard Clode called the $78 billion guidance “well ahead of even the most bullish expectations” and noted it marks the fourth straight quarter of accelerating growth. At a P/E of 48.5x — steep, but not insane for a company growing revenue 73% annually — the valuation math still works if the AI spending cycle holds.

The real takeaway isn’t that Nvidia is broken. It’s that the stock has become a barometer for the entire AI trade, and right now, that trade is running on emotion, not logic. When the best earnings report in the company’s history triggers a $260 billion wipeout, you’re not trading fundamentals anymore — you’re trading sentiment. And sentiment, unlike Nvidia’s revenue growth, can turn on a dime.

Article

American Investors Are Quietly Dumping U.S. Stocks at Record Pace

Something strange is happening on Wall Street. While headlines fixate on Nvidia earnings and AI hype cycles, American investors are doing something they haven’t done at this scale in over 16 years: pulling their money out of U.S. stocks and shipping it overseas.

The numbers are jarring. In just the first eight weeks of 2026, U.S.-domiciled investors yanked $52 billion from domestic equity products — the largest outflow for that period since at least 2010. Zoom out six months and the total hits $75 billion, according to LSEG/Lipper data. This isn’t a trickle. It’s a full-blown rotation.

Where’s the money going? Everywhere that isn’t the S&P 500, apparently. Bank of America’s February fund manager survey showed investors switching from U.S. to emerging market equities at the fastest clip in five years. South Korea led the way with $2.8 billion in inflows, followed by Brazil at $1.2 billion. In total, $26 billion has flowed into emerging market equities since January.

The performance gap tells the story. Over the last 12 months, the S&P 500 has returned about 14%. Decent in a vacuum — but Seoul’s KOSPI has doubled. Tokyo’s Nikkei is up 43% in dollar terms. Europe’s STOXX 600 surged 26%. Even Shanghai’s CSI 300 posted 23% returns. If you were parked exclusively in U.S. equities, you missed the party happening everywhere else.

Valuations add fuel to the fire. The S&P 500 still trades at roughly 21.8 times forward earnings. Europe? Around 15 times. Japan sits at 17. China is a bargain-basement 13.5 times. When UBS’s head of European equity strategy says his U.S. wealth clients are “all talking about investing more offshore,” you know the narrative is shifting fast.

The weakening dollar — down about 10% against a basket of currencies since January 2025 — makes this rotation even more interesting. Yes, it’s more expensive to buy foreign assets. But the returns on those assets, converted back to dollars, get a nice tailwind. European bank stocks alone ripped 67% higher last year and are still climbing in 2026.

Here’s the kicker: since Trump’s inauguration, U.S. investors have poured nearly $7 billion into European equity products. During his entire first term from 2017 to 2021, they pulled out $17 billion. That’s a complete reversal in sentiment.

Does this mean you should dump your U.S. holdings tomorrow? Not necessarily. But the “buy America” trade that worked almost effortlessly since 2009 is showing cracks. The smart money isn’t panicking — it’s diversifying. And when the biggest rotation in 16 years is underway, the worst thing any investor can do is pretend it isn’t happening.

Article

NVIDIA Reports Tonight and the Entire AI Trade Hangs in the Balance

NVIDIA drops its fiscal Q4 2026 earnings after the bell today, and it’s not an exaggeration to say this single report could set the tone for tech stocks for the rest of the year. Wall Street expects $65.7 billion in revenue — a 67% jump from a year ago — with earnings per share around $1.53, up roughly 72%. Those are staggering numbers for any company. For NVIDIA, it’s just another Tuesday.

The real story isn’t whether NVIDIA beats estimates. Prediction markets are pricing in a 94.5% probability it clears the $1.52 EPS bar, and the company has beaten for twelve straight quarters. The question that actually matters: how big is the Blackwell supercycle, and is the best yet to come?

Here’s what makes this quarter different. CEO Jensen Huang has described demand for NVIDIA’s next-generation Blackwell chips as “off the charts” and “insane,” with clouds “sold out” and GPU capacity fully utilized across Blackwell, Hopper, and Ampere generations. The company has $350 billion in Blackwell and Rubin pipeline through the end of calendar 2026. Meanwhile, every major hyperscaler is throwing money at AI infrastructure at a pace that’s shocked even the bulls — Meta guided $115 billion to $135 billion in capex, Alphabet $175 billion to $185 billion, and Amazon a jaw-dropping $200 billion. Every one of those dollars flows upstream to NVIDIA’s top line.

Gross margins will be closely watched. NVIDIA is targeting 75% non-GAAP for the quarter, up from 73.6% last quarter. Holding margins in the mid-70s while ramping Blackwell production at scale would silence bears who’ve argued that rising input costs and supply chain complexity will squeeze profitability. CFO Colette Kress has committed to maintaining this margin profile into fiscal 2027 — a pledge investors will want to see backed up with numbers.

Then there’s the China wildcard. NVIDIA has guided zero data center compute revenue from China for this quarter, after H20 chip sales came in at just $50 million in Q3. Any thaw in export restrictions or traction with workaround products would represent pure upside that’s not baked into a single analyst model.

But the real fireworks will come from guidance. Wall Street expects NVIDIA to guide Q1 fiscal 2027 to roughly $71 billion in revenue. If the company surprises above $75 billion — which some analysts believe is possible given the hyperscaler spending surge — expect a wave of upward revisions across the street. Loop Capital Markets already models $9.56 in fiscal 2027 earnings, well above the consensus $7.76. If tonight’s guidance supports that kind of trajectory, the stock could break out of the $180-range it’s been stuck in.

One note of caution: Fundstrat’s Mark Newton has warned about a potential “false breakout” if the numbers merely meet expectations without a blowout guide. After a 47% run year-to-date, good-not-great may not be good enough.

For investors, the playbook is straightforward. This isn’t just an NVIDIA earnings call — it’s a referendum on the AI capital cycle itself. Strong numbers and a raised outlook validate every hyperscaler’s spending plan and lift the entire AI ecosystem, from server makers like Super Micro and Dell to chip foundries like TSMC. A disappointment would send tremors through a sector that’s priced for perfection. Either way, this is one report you don’t want to miss.

Article

AI Just Wiped Billions Off IBM — And Wall Street Thinks It’s an Overreaction

IBM just had its worst single-day drop in over 25 years — a 13.15% plunge that sent shares to $223 and erased roughly $30 billion in market value. The culprit wasn’t bad earnings, a CEO scandal, or a geopolitical crisis. It was a software tool.

Anthropic announced that its Claude Code tool can now automate the modernization of COBOL — the ancient programming language that still powers roughly 95% of ATM transactions and the core infrastructure of nearly every major bank and insurance company on the planet. IBM has built an empire around maintaining, consulting on, and housing this legacy code on its mainframes. If AI can do that job automatically, two of IBM’s most stable and lucrative revenue streams — legacy modernization consulting and mainframe infrastructure — face serious questions.

Month-to-date, IBM is now down about 27%, putting it on pace for its worst month since 1992. The market’s message was blunt: if COBOL modernization gets automated, IBM’s consulting gravy train could derail fast.

But here’s where it gets interesting. Wall Street doesn’t agree with the panic. The average analyst price target still sits at $327, implying nearly 50% upside from current levels. And IBM isn’t exactly a sitting duck — the company already has its own Watsonx Code Assistant for Z, specifically targeting COBOL modernization. In other words, IBM saw this coming and has been quietly building its own answer.

The fundamentals paint a more nuanced picture than the stock chart suggests. Revenue grew 4.5% over the last 12 months to $65 billion. The most recent quarter posted 9.1% year-over-year growth — actually outpacing the S&P 500’s 7.5%. Operating cash flow margins sit around 20.6%, and the company holds $15 billion in cash. At today’s beaten-down price, IBM trades at a price-to-free-cash-flow ratio of 17.8 versus the S&P 500’s 21.7 — making it technically cheaper than the broader market.

The real question for investors isn’t whether AI threatens COBOL maintenance revenue — it clearly does. The question is whether that threat justifies a 27% drawdown in a company generating $65 billion in revenue, paying a healthy dividend, and pivoting hard toward hybrid cloud and AI. History offers some reassurance: during the 2022 sell-off, IBM fell 20% and recovered by November. During COVID, it dropped 39% and bounced back by late 2022.

One AI tool didn’t kill the mainframe. But it did give investors a rare chance to reassess Big Blue at prices not seen in years — with nearly every analyst on Wall Street saying the market got this one wrong.

Article

The Supreme Court Killed Tariffs and Left a $175 Billion Mess

On Friday, the Supreme Court did something Wall Street had been anticipating for months — it struck down President Trump’s sweeping IEEPA tariffs in a decisive 6-3 ruling, declaring the president had overstepped his emergency powers. Markets popped immediately. The S&P 500 climbed 0.7%, the Nasdaq surged 0.9% to snap a brutal five-week losing streak, and for a brief moment, the trade war felt like it was actually over.

It wasn’t. Before the ink was dry on the ruling, the White House announced a fresh 10% tariff under the Trade Act of 1974 — a completely different legal authority the Court didn’t touch. Over the weekend, reports surfaced that the administration is already mulling an increase from 10% to 15%. Futures opened lower Monday, with the S&P 500 sliding about 0.5%, reminding everyone that tariff whiplash remains the defining feature of this market cycle.

But here’s where it gets interesting for traders. The ruling triggered a massive refund question that nobody on Wall Street can agree on. Morgan Stanley pegs the potential refund at roughly $85 billion. RSM’s chief economist says $100 to $130 billion. Raymond James went all the way to $175 billion, matching a University of Pennsylvania model. That’s real money flowing back to importers — and retailers in particular. The National Retail Federation is already calling it an “economic boost,” urging the lower courts to expedite refunds so companies can “reinvest in their operations, their employees, and their customers.” Justice Kavanaugh, though, offered a more candid assessment of the refund process: a “mess.”

The timing of this ruling couldn’t be more relevant for the week ahead. Home Depot reports earnings Tuesday, followed by Lowe’s and TJX on Wednesday. These are exactly the tariff-sensitive retail names that stand to benefit most from potential refunds and the removal of IEEPA duties. Analysts at RSM flagged “enormous potential winners” in the retail and manufacturing sectors. The question for each of these earnings calls will be straightforward: how much did tariffs actually cost you, and what’s the refund math look like?

Meanwhile, the macro backdrop is complicated. The PCE inflation report that dropped Friday showed core prices running at a 3% annual rate — hotter than expected and well above the Fed’s 2% target. Q4 GDP came in at a sluggish 1.4%, though the longest government shutdown in history deserves most of the blame there. The Fed is now expected to hold rates steady until at least July, with markets pricing in two cuts for 2026 and roughly 40% odds of a third. Losing the IEEPA tariffs removes about half a percentage point of inflationary pressure, according to central bank estimates — a modest tailwind, but not the game-changer bulls were hoping for.

The bottom line: tariffs got killed by the Supreme Court and immediately came back as a zombie under different legal authority. The refund bonanza could be enormous — or could get tied up in courts for years. Retail earnings this week will be the first real-time test of who benefits and who’s still stuck paying import taxes. Keep your eyes on those calls. The market just entered a new chapter of trade policy uncertainty, and the winners are the companies that hedged smartly and the traders who stay nimble.

Article

Supreme Court Torched Trump’s Tariffs — He Raised Them Anyway

In what might be the most dramatic 24 hours in trade policy history, the Supreme Court struck down a massive chunk of President Trump’s tariff agenda on Friday — and Trump responded by making them even bigger.

The 6-3 ruling found that Trump had overstepped his authority by using the International Emergency Economic Powers Act (IEEPA) to impose tariffs, effectively torching roughly 60% of his existing trade levies. Justices Gorsuch and Barrett sided with the majority, prompting Trump to call the decision “ridiculous, poorly written, and extraordinarily anti-American.” Within hours, he slapped a 10% global tariff under Section 122 of the Trade Act of 1974 — a different legal tool that allows temporary levies for 150 days without Congressional approval. By Saturday morning, he’d already bumped it to 15%, “effective immediately.”

For investors, the whiplash is the point. Markets rallied Friday on the initial ruling, with stocks glossing over weak GDP data (Q4 came in at just 1.4% annualized) and a core inflation reading that held stubbornly at 3%. The hope? That dismantling the IEEPA tariffs could mean lower prices, potential corporate refunds, and one fewer headwind for the Fed to worry about. Estimates on what the government might owe in refunds range from $85 billion (Morgan Stanley) to a staggering $175 billion (Raymond James and a University of Pennsylvania model). That’s real money flowing back to importers — and potentially, to their shareholders.

But here’s the wrinkle nobody should ignore: the tariffs aren’t actually going away. Trump is simply rerouting them through a different legal channel. The 15% global levy is real, and the administration has signaled more “legally permissible” tariffs are coming in the months ahead. TD Cowen’s Chris Krueger expects 2026’s tariff strategy to be “all gas, some temporary brakes.”

The Fed is watching closely. Rate-cut expectations barely budged after the ruling — traders still expect two cuts this year, with July now looking more likely than June for the next move. The court decision removes one inflationary pressure, but Trump’s counterpunch adds a new one. It’s a wash, for now.

What should investors do heading into Monday? First, watch the State of the Union address on Tuesday — Trump is likely to use it as a tariff megaphone. Second, keep an eye on retail and manufacturing stocks, which RSM’s chief economist flagged as “enormous potential winners” from the refund scenario. Third, don’t assume the volatility is over. The legal battle is just shifting venues, and Trump has made it clear he’ll use every tool in the toolbox to keep tariffs alive. The only thing that’s certain is that trade policy will keep making headlines — and moving markets — for months to come.

Article

Supreme Court Killed the Tariffs — But the Market Party May Be Short-Lived

In a ruling that sent shockwaves through Wall Street on Friday, the Supreme Court struck down President Trump’s sweeping tariffs in a decisive 6-3 decision. The justices determined Trump exceeded his authority when he invoked emergency powers under IEEPA to slap import taxes on most of America’s trading partners. Just like that, the signature economic policy of this administration got its legs cut off by the highest court in the land.

Markets popped on the news. The S&P 500 gained 0.7% to close at 6,909, the Dow added 231 points, and the Nasdaq climbed 0.9% — snapping its worst weekly losing streak since May 2022. Five straight weeks of red, erased in a single Friday afternoon. The biggest winners? Companies with heavy import exposure. Jefferies immediately named Nike, Yeti, and SharkNinja as top beneficiaries, while Temu parent Pinduoduo surged over 4.5% as traders priced in cheaper Chinese goods flowing back into the U.S. market.

But here’s what should keep you sharp: the underlying economic picture is deteriorating fast. The same day the tariffs fell, we learned the U.S. economy grew at a measly 1.4% annualized rate in Q4 — a full percentage point below estimates and a nosedive from 4.4% in Q3. Meanwhile, the PCE inflation gauge (the Fed’s favorite) came in hotter than expected at 2.9% year-over-year, with core inflation ticking up to 3.0%. Slow growth plus sticky inflation is the textbook definition of a word nobody wants to hear: stagflation.

And the so-called “smart money” isn’t celebrating. According to VandaTrack, retail investors barely flinched after the ruling. This week is tracking as one of the weakest for net retail inflows in recent years. The mom-and-pop crowd that powered last year’s rally? They’re sitting on the sidelines, watching.

The tariff reversal also doesn’t exist in a vacuum. Oil prices surged 6% this week as the U.S. built up its largest military presence in the Middle East since 2003, eyeing a potential conflict with Iran that could choke 20% of global oil supply through the Strait of Hormuz. Gold blasted to $5,125 an ounce, and silver surged 9% to $84.50 — classic fear trades that tell you the big money is hedging, not going all-in on equities.

Meanwhile, Kalshi prediction market contracts put a 58% probability on the S&P 500 falling to at least 6,200 at some point this year — an 11% correction from its record high. And history backs it up: in midterm election years with a new president, the median intra-year drawdown for the S&P 500 is a gut-punching 21%. That puts the odds of a full-blown bear market at roughly 50% in 2026.

The bottom line? The tariff ruling was a genuine win for consumers and importers — smaller businesses especially, who lacked the supply chain muscle to dodge the duties. But one court decision doesn’t fix weak GDP, stubborn inflation, geopolitical tinderboxes, or a market trading at 21.5x forward earnings when the historical average is 20x. The relief rally feels good, but the fundamentals are whispering something different. Stay alert, stay diversified, and don’t mistake one green day for an all-clear signal.

Article

Bitcoin’s Dirty Secret: It Trades Like a Tech Stock, Not Gold

Bitcoin just fell over 50% from its October peak of $126,000 to a $60,000 bottom. It’s now hovering around $68,000. And if you bought it thinking you owned “digital gold,” you might want to sit down for this.

Here’s the uncomfortable truth the crypto faithful don’t want to hear: Bitcoin doesn’t behave like a hedge. It behaves like a tech stock. Specifically, it now trades almost identically to the Nasdaq.

Data from InvestorPlace’s Eric Fry shows that Bitcoin’s 90-day price correlation with the Nasdaq Composite has risen so dramatically over the past two years that it’s now nearly as high as Nvidia’s correlation with the same index. Read that again. A cryptocurrency with zero connection to the tech sector now moves in near-lockstep with a chip maker that accounts for the largest single weight in the Nasdaq.

Over the past six months, Bitcoin has shown a tighter correlation with the Nasdaq than Tesla, Microsoft, Meta, or Apple. Fry quips that Bitcoin might as well be called the “Magnificent Eight.” It’s a funny line — until you realize what it means for anyone treating BTC as portfolio insurance.

The math problem is simple. Bitcoin’s total market cap sits around $1.8 trillion. That sounds big until you compare it to U.S. Treasuries ($27 trillion) and investable gold (~$15 trillion). Combined, the two traditional safe havens offer more than $40 trillion in tradable assets. If Wall Street’s $73 trillion stock market enters full panic mode, Bitcoin’s relatively tiny pool can’t absorb the rush. And based on recent behavior, it wouldn’t even try — it’d be selling off alongside everything else.

The current SaaS meltdown has proven the point in real-time. As software stocks cratered, Bitcoin followed them down. Not because of any crypto-specific catalyst, but because that’s what correlated assets do.

None of this means Bitcoin is a bad investment. Over the long haul, it will likely continue gaining global adoption. But calling it “digital gold” is like calling a motorcycle a tank because they both have engines. The function is completely different.

For investors looking for actual downside protection, the old playbook still works: Treasuries, gold, and cash-flowing businesses that sell things humans will always need — food, energy, healthcare. These won’t give you 10x returns in a bull run, but they also won’t halve your money when the Nasdaq catches a cold.

The takeaway? Own Bitcoin if you want. But know what you actually own: a high-volatility risk asset that rides the same wave as tech stocks. If you’re counting on it to save your portfolio when things get ugly, you might be the one who needs saving.

Article

Meta Just Locked In Millions of NVIDIA Chips — Here’s Why It Matters

Meta just made one of the biggest bets in AI hardware history — and it tells you everything you need to know about where the real money in artificial intelligence is heading.

On Tuesday, Meta and NVIDIA announced a “multigenerational” partnership that will see Meta build hyperscale data centers powered by millions of NVIDIA’s Blackwell and next-generation Rubin GPUs. But here’s the part most people are missing: Meta isn’t just buying graphics processors. For the first time, a major tech giant is making a large-scale purchase of NVIDIA’s Grace CPU as a standalone chip. That’s NVIDIA muscling directly into Intel and AMD’s turf — and it signals a fundamental shift in how AI infrastructure gets built.

The numbers behind this deal are staggering. Meta previously purchased around 350,000 H100 chips by the end of 2024 and scaled to 1.3 million GPUs by the end of 2025. Now it’s ordering “millions” of next-gen chips. Meanwhile, Meta has guided AI infrastructure spending of $115 billion to $135 billion for 2026 — nearly double last year’s $72.2 billion. That’s not a company hedging its bets. That’s a company going all-in.

Why the sudden hunger for CPUs alongside GPUs? One word: inference. As AI shifts from training massive models to actually running them at scale — think agentic AI, real-time recommendations for 3 billion users, AI features inside WhatsApp — CPUs become critical. They handle the general-purpose computing tasks that feed data to and from the GPUs. Without enough CPU horsepower, those expensive GPUs sit idle. Analysts at Semianalysis recently noted that one of Microsoft’s OpenAI data centers now requires “tens of thousands of CPUs” just to process the petabytes of data generated by its GPUs.

For investors, this deal has several implications worth watching. First, it’s a massive validation for NVIDIA ahead of its February 25th earnings report. The stock popped 1.6% on Wednesday and has been one of the market’s primary sentiment drivers. Second, this “one-throat-to-choke” approach — where Meta sources GPUs, CPUs, and networking equipment from a single vendor — could pressure competitors like AMD, Intel, and even Google’s TPU business. Third, it suggests the AI capex boom isn’t slowing down. If anything, it’s accelerating as companies move from the training phase to deploying AI at consumer scale.

Mark Zuckerberg himself framed the deal in unmistakable terms, saying Meta will use NVIDIA’s Vera Rubin platform to “deliver personal superintelligence to everyone in the world.” Whether or not you buy the superintelligence timeline, the spending is real, the chips are real, and the competitive implications across the semiconductor industry are enormous. With NVIDIA earnings just days away, this partnership just raised the stakes — and the expectations — for what Jensen Huang delivers next.