Date: August 17, 2026

The market is still paying up for one thing above all: AI-linked growth with visible demand and hard capacity behind it. Software has rallied, but the leadership has been narrow, and investors have shown less patience for stories that cannot translate into revenue soon.

The cleanest signal sits in the infrastructure layer. When AI demand accelerates, the bottlenecks are not ideas. They are compute supply, power, networking, and the companies that can deliver them at scale.

The headline driving today’s issue is striking: Axios, citing Bloomberg figures, reported Anthropic’s annualized revenue run rate has moved above $65 billion ahead of an expected IPO process already underway. Anthropic also disclosed earlier this summer that it confidentially submitted a draft S-1 to the SEC on June 1, 2026.

That headline matters, but investors face a practical problem. You cannot buy Anthropic today. So the trade becomes: who captures the most durable economics from that growth curve while remaining investable now?

• Public AI platform winners (benefit from usage, but face pricing pressure and model churn).
• Hyperscalers (own distribution, but AI margins can be diluted by capex cycles).
• Semis and infrastructure (benefit directly from capacity buildout and recurring upgrade cycles).

Today’s single best stock expression, in our view, is the infrastructure toll collector with the most obvious beneficiary profile as Anthropic, OpenAI, and the hyperscalers fight for capacity.

We want evidence that serious capital is still flowing into the AI buildout, not just into the most popular tickers. One of the clearest tells is when frontier labs sign multi-year capacity agreements that show up in public-company filings.

On April 6, 2026, Broadcom disclosed that Broadcom, Google, and Anthropic expanded their strategic collaboration, with Anthropic set to access approximately 3.5 gigawatts of next-generation TPU-based compute beginning in 2027. That is not a small pilot. It is a scale commitment.

What the market expects: Investors already expect AI infrastructure spending to remain elevated. Many AI-linked stocks are priced for strong growth, leaving less room for execution mistakes.

What may not be fully priced: The pace and duration of TPU and ASIC buildouts tied to named customers and explicit gigawatt-scale capacity plans. Broadcom’s role as an enabling partner to Google’s TPU supply chain creates a different demand stream than the standard “GPU cycle” framing.

What could exceed expectations: If Anthropic’s reported revenue trajectory is even directionally accurate, the compute arms race has to widen, not slow. The more frontier labs chase enterprise workloads, the more they need predictable unit economics and reliable capacity. That favors industrialized supply chains.

What could disappoint: A sharp slowdown in enterprise AI spend, model price compression that hits usage growth, or a shift in mix away from TPU/ASIC supply chains where Broadcom participates. Another risk is timing. Agreements beginning in 2027 can create air pockets if investors get impatient in the interim.

• Primary risks: valuation risk, AI capex digestion cycles, customer concentration, competitive dynamics in custom silicon and networking, and execution risk in delivering capacity at scale.
• Thesis invalidation: clear evidence that the frontier labs are materially reducing capacity commitments, or that the economic capture is shifting decisively away from Broadcom’s portion of the stack.

Featured stock: Broadcom (AVGO).

If you believe the Anthropic headline, you should also believe the constraint is not demand. It is supply. Broadcom sits in the part of the value chain that scales when the frontier labs need real capacity delivered, not slide decks.

Why This Stock Now

Anthropic’s reported move to a revenue run rate above $65 billion ahead of an IPO is a reminder that AI is no longer “future optionality” for a handful of companies. It is becoming a present-tense revenue engine.

You cannot buy Anthropic today. You can buy the infrastructure beneficiary that has already put customer-linked, gigawatt-scale capacity agreements on the record.

The Business

Broadcom is a scale semiconductor and infrastructure software company. For this thesis, the key point is its role in enabling high-volume, specialized compute supply chains, including TPU-based capacity delivered through partners.

In AI, the “product” is not only chips. It is the ability to deliver compute at the required performance per watt, with supply reliability, and with an ecosystem that customers can build on for years.

Why Wall Street Is Paying Attention

The attention is rational. The biggest buyers of AI compute have started to look like industrial planners, not experimental labs. When Broadcom’s April 6 disclosure points to Anthropic accessing roughly 3.5 gigawatts of next-generation TPU-based compute starting in 2027, it frames demand in the language public markets respect: capacity, timing, and scale.

Meanwhile, Anthropic’s own IPO process is not hypothetical. The company announced it confidentially submitted a draft S-1 to the SEC on June 1, 2026. If the IPO window opens, public investors will be forced to assign real multiples to AI revenues and to the infrastructure that supports them.

What’s Driving the Opportunity

• AI demand is turning into measured revenue: Bloomberg figures cited by Axios suggest Anthropic’s annualized revenue run rate has moved above $65 billion.
• Capacity commitments are scaling: Broadcom’s filing describes expanded collaboration with Google and Anthropic that includes access to approximately 3.5 gigawatts of next-generation TPU-based compute beginning in 2027.
• Second-order effect: As AI vendors chase enterprise workloads, reliability and cost per unit matter more. That tends to pull spend toward standardized, scaled supply chains rather than one-off experimentation.

What Could Go Wrong

The core risk is that markets are already crowded into “AI winners,” and any wobble in capex, a pause in enterprise spend, or a shift in customer architecture can hit the group together.

There is also a timing risk. The 2027 start date on the 3.5 gigawatt disclosure is a double-edged sword. It supports the durability of demand, but it can tempt investors to discount the benefit too aggressively if near-term numbers do not move fast enough.

Finally, investors should respect competitive risk in custom silicon supply chains. Winning one cycle does not guarantee the next. The thesis depends on Broadcom continuing to be involved in the scaled delivery of next-generation compute capacity.

The Bottom Line

Anthropic’s reported $65 billion revenue run rate is an attention-grabbing number, but the actionable investment takeaway is simpler: if AI revenue is compounding at that speed, the scarce asset is compute capacity and the industrial supply chain behind it.

That is why Broadcom is our one-stock pick today. It is a public, scalable way to express the next leg of the AI buildout, anchored by disclosed collaboration tied directly to Anthropic and Google’s TPU capacity roadmap.

Action: Treat AVGO as the “infrastructure expression” of the Anthropic IPO cycle. If you want the upside of AI revenue acceleration without waiting on an IPO allocation, this is a high-conviction place to focus.

— Trading Stocks Now