FEATURED

WD-40’s Best Quarter in Years

Hey there, bargain hunter.

Something unusual happened this week. Meta Platforms, a stock that spent most of 2026 getting punished for spending too much, is suddenly getting rewarded for spending even more.

Meta shares rallied on Friday, lifting their gains for the week to 15%, on pace for the best weekly performance since early 2024, as optimism builds around CEO Mark Zuckerberg’s artificial intelligence strategy. From the mid-$540s in late June to roughly $673 intraday on July 10, the stock has made back nearly everything it lost in 2026 in less than three weeks. With the latest rally, shares have erased most of their year-to-date losses and are now slightly above flat for the year.

That is not a momentum glitch. Three separate things happened in rapid succession. Each one is meaningful on its own. Together, they rewrote the investment case.

What Actually Happened

Start with the chip. At the center of this week’s move is a custom AI processor called Iris, with mass production scheduled to begin in September 2026. It is Meta’s fourth-generation chip under its Meta Training and Inference Accelerators program, known as MTIA, designed in-house with Broadcom and manufactured by TSMC. According to an internal memo reviewed by Reuters, testing the chip took only six weeks and found no major issues. That is notably fast for a data center-grade AI chip, where post-design validation typically takes much longer. Industry observers called it a meaningful milestone for a program that has struggled since its launch more than five years ago.

Slight tangent, but it matters: custom silicon is how Google’s AI economics became unbeatable. TPUs let Google run inference at a fraction of what a straight Nvidia GPU cluster costs. Meta just reached the same milestone Google hit years ago.

Meta’s capex guidance for 2026 sits at $125 billion to $145 billion, mostly going into data centers, GPUs, and custom silicon. The company is targeting 7 gigawatts of computing capacity in 2026, then doubling that to 14 gigawatts in 2027. Iris is central to hitting those targets without the cost structure blowing out further. Bank of America analyst Justin Post noted that an internal memo suggests Meta could be building capacity at a materially lower cost per gigawatt than the Street previously estimated, with prior estimates around $45 billion per gigawatt potentially coming in much lower. If that efficiency holds, the 2027 cost story gets significantly more interesting.

Then there is the cloud business.

Meta is building a cloud infrastructure business called Meta Compute, which will sell access to AI computing power and models directly to outside companies. That puts Meta in the same conversation as AWS, Microsoft Azure, and Google Cloud. Bloomberg reported July 1 that the service would offer two products: developer access to AI models hosted on Meta’s own GPU infrastructure, similar to how AWS Bedrock works, and raw GPU capacity rented to third parties by the hour, similar to what CoreWeave offers. The announcement drove Meta shares up roughly 8 to 10% on that single day.

Here is the uncomfortable part for current Meta partners.

Meta currently has a $21 billion commitment with CoreWeave, signed in March 2026, as part of CoreWeave’s $99.4 billion total revenue backlog. Meta is CoreWeave’s second-largest customer. Meta becoming a compute landlord would flip that relationship over time. Shares of neocloud companies CoreWeave and Nebius Group both dropped sharply following the Meta Compute news. That reaction tells you everything about how seriously the market is taking this pivot.

The Model Launches Nobody Expected This Week

Three months after introducing Muse Spark, its first proprietary foundation AI model, Meta made two more significant announcements this week. On Tuesday, Meta released Muse Image, a new AI model for creating images aimed at creators and advertisers. On Thursday, the company unveiled Muse Spark 1.1, aimed at running agentic and coding workloads, putting it directly against similar products from OpenAI and Anthropic.

The Muse Spark 1.1 launch is being read as one of the clearest signals yet that Meta is starting to convert its heavy AI spending into concrete commercial products. That perception matters more than the model spec sheet right now. Bank of America expects Q2 revenue to give investors more confidence that Muse Spark is beginning to move the advertising needle. Meta also committed approximately $10 billion (C$13 billion) to a new AI data center in Alberta, Canada, adding another concrete infrastructure milestone to the week.

The BofA Move That Changed the Week

On July 9, Bank of America added Meta to its US 1 List, the bank’s designation for its highest-conviction buy-rated ideas. It is a meaningful step up from a standard buy rating. BofA analyst Justin Post reiterated a Buy rating and an $835 price target, pointing to the internal chip memo as evidence Meta could expand computing capacity more efficiently than previously estimated.

The broader BofA thesis: Meta trades at roughly 18 times revised 2027 earnings estimates, below its historical average of 21 times. The bank projects Meta will generate $254.6 billion in revenue in 2026 and $311.3 billion in 2027. Their view is that Meta’s core advertising business is strong enough to absorb the spending while AI improvements compound on top of it. Post and his co-analyst pushed back on the capex-fear read, arguing the infrastructure is being built either way, and the option to monetize it externally through Meta Compute is upside the current stock price does not yet reflect.

The Numbers Behind the Move

Q1 2026 results: total revenue of $56.3 billion, up 33% year over year. Net income of $26.8 billion. Gross margin around 81.9% and EBIT margin near 42%. Operating cash flow near $32.2 billion, free cash flow around $13.2 billion after roughly $19 billion in capital spending. Those are not the margins of a company struggling to survive its spending cycle.

Meta guided Q2 2026 revenue at $58 billion to $61 billion, implying roughly 25% year-over-year growth and above the analyst consensus heading into that guidance. Wall Street currently expects Q2 EPS of around $7.18 on revenue of roughly $60.2 billion, with earnings scheduled for July 29.

The cloud business thesis is longer-dated. Building and profiting from an enterprise cloud takes years. Amazon and Google both spent heavily before AWS and Google Cloud hit profitability. Meta would be starting from scratch in that market, even with a massive computing base to monetize.

Who Benefits. Who Gets Hurt.

Applied Materials climbed nearly 7% on Thursday after the Iris chip headlines. Lam Research and KLA also posted solid gains as investors priced in new demand for the wafer fabrication equipment those companies make. When a company the size of Meta moves to proprietary chip designs, it creates real, concrete demand for the semiconductor equipment supply chain. Broadcom and TSMC benefit directly from new Iris production volume. Micron Technology, AMD, and Broadcom all gained on the session as well.

CoreWeave is the clearest loser in this story. The company posted a net loss of $740 million in Q1 2026, carries roughly $24.9 billion in debt, and has capex guidance of $31 billion to $35 billion for the full year 2026. The bull case for CoreWeave rests almost entirely on $99.4 billion in contracted revenue backlog, a large chunk of which comes from Meta. If Meta starts building its own compute and reducing its reliance on third-party neocloud providers over time, that backlog durability becomes a real question.

The Part That Should Give You Pause

Here is what nobody is talking about loudly enough.

Four U.S. states, California, Colorado, Kentucky, and New Jersey, are seeking approximately $1.4 trillion in penalties in a youth-safety trial scheduled for August 2026 in Oakland, California. Meta disclosed the figure in a court filing, noting the number is close to its entire market capitalization of roughly $1.5 trillion. Meta has contested the figure vigorously, calling it without precedent in the history of consumer protection enforcement. The company argues the states calculated penalties by multiplying every qualifying teen user by the maximum fine under each applicable law, regardless of actual harm. Even so, a trial is a trial, and August is right around the corner from July 29 earnings.

And then, on July 10, the European Commission issued preliminary findings that Instagram and Facebook breached the Digital Services Act through addictive design features like infinite scroll, autoplay, and highly personalized recommendations. If confirmed, Meta could face fines of up to 6% of its global annual revenue, which analysts estimate could reach nearly $12 billion. The findings are not final and Meta has contested them. But that is a material legal overhang landing on the same day the stock is trying to make new weekly highs.

The market is focused on the cloud thesis and the chip timeline. The regulatory exposure is real and not going away by July 29.

What to Watch Before Earnings

• Does Meta formalize a launch date, pricing structure, or customer pipeline for Meta Compute on the Q2 call?
• Does Iris production begin on schedule in September? Any slip in timing changes the 2027 cost story materially.
• What is management’s actual language around cloud revenue: is this an adjacency or a strategic priority?
• Does the EU fine number firm up, and does Meta signal a settlement posture or a fight?
• Wolfe Research models roughly 20% EPS upside per gigawatt of cloud capacity monetized at a $25 billion rate, backing META with an $800 target despite sharply higher capex. Watch whether other analysts start adopting that framework.
• CoreWeave’s next earnings: any revenue guidance cuts tied to Meta contract risk would confirm the worst-case read.
• BNP Paribas expects Meta may raise its full-year capex guidance again on the Q2 call, to a range of $135 billion to $155 billion. If that number shows up, watch how the stock reacts this time versus April.

If Meta can build capacity at half the cost previously modeled, the economics of an in-house AI cloud become credible. That is still a big if. The July 29 call is where the if starts becoming a yes or no.

The market has shifted its read on Meta’s spending from liability to asset in less than three weeks. Whether that shift holds depends entirely on what Zuckerberg says when the cameras are on.