브로드컴은 AI 매출이 3배 증가한 167억 달러를 기록했으며, 이는 고객 수요가 현재 동사의 2,300억 달러 규모 자본 지출 로드맵을 상회하고 있음을 시사한다.

Broadcom’s fiscal third quarter generated $16.7 billion in AI semiconductor revenue—more than triple the figure from the same period last year—and delivered a critical detail the headline number alone obscures: CEO Hock Tan told investors Wednesday that actual customer demand already surpasses even Broadcom’s own $115 billion fiscal 2027 target. According to the official Q3 earnings press release, the company’s forecast is intentionally conservative because fabrication capacity and advanced packaging supply, not orders, remain the binding constraint.

This disclosure flips the narrative that has dominated Broadcom coverage throughout the first half of 2026. The prevailing risk was whether demand would hold. It is holding—and exceeding expectations.

**Anthropic Is Now Broadcom's Biggest Customer in Waiting**

The most structurally significant takeaway from Wednesday’s earnings call was not the striking $230 billion fiscal 2028 target, but rather that Anthropic is on track to become Broadcom’s largest XPU customer in 2027, surpassing Google, and will retain that position through 2028, per the Benzinga Q3 2026 earnings call transcript. Tan noted that Anthropic will deploy an additional 5 gigawatts of next-generation TPU v8i capacity in 2027, with Broadcom maintaining clear visibility to deliver another 10 gigawatts in 2028. This progression—one gigawatt of Ironwood in 2026, 5 gigawatts of v8i in 2027, and 10 more gigawatts in 2028—represents one of the most aggressive compute capacity ramps any AI laboratory has publicly committed to.

The shift carries editorial weight because Broadcom’s AI narrative has been built quarter by quarter around its relationship with Google. Google’s TPU programs served as the revenue anchor, and every analyst model questioning the achievability of the $56 billion fiscal 2026 forecast primarily hinged on Google. Anthropic’s emergence as the new anchor customer in 2027 reduces Broadcom’s exposure to Google concentration risk—the very concern that prompted Macquarie to downgrade Broadcom to Neutral from Outperform in June. That downgrade triggered a sharp stock decline and has weighed on Broadcom’s performance relative to peers ever since.

OpenAI, whose first-generation Jalapeño accelerator Broadcom shipped during the third quarter, is positioned to rank as the second-largest XPU customer in 2027, according to the Benzinga call transcript. Tan stated that Jalapeño outperforms Nvidia’s Grace Blackwell GPU for inference workloads, placing the custom chip at performance parity with or above Nvidia’s latest merchant silicon for the specific tasks OpenAI requires. Broadcom and OpenAI are already developing a second-generation chip and outlining a third. For Meta, Broadcom expects to begin production shipments of its latest MTIA custom accelerator in the fourth quarter, optimized for large-scale inference and recommendation workloads. “Four of our six customers will grow to enormous scale,” Tan said, per the Q3 2026 Investing.com transcript.

**Q3 by the Numbers**

Total consolidated revenue for the quarter ended August 2 reached $29.59 billion, an 86% year-over-year increase that surpassed the Wall Street consensus of approximately $29.36 billion, per the official earnings press release. The result extended Broadcom’s streak of adjusted EPS beats to nine consecutive quarters, with non-GAAP diluted EPS of $3.32 clearing the consensus range of $3.22 to $3.24.

Profitability metrics highlighted the operating leverage inherent in custom silicon at scale. Non-GAAP operating income grew 92% year over year to $20.1 billion, yielding an operating margin of 68%. This rate underscores how minimal incremental overhead the AI chip business requires relative to the revenue it generates, per the earnings press release. Free cash flow hit a record $13.7 billion, representing 46% of quarterly revenue. The company concluded the period with $24.0 billion in cash after paying down $5.6 billion in debt and returning $3.1 billion in dividends.

AI semiconductor revenue of $16.7 billion now constitutes 56% of Broadcom’s total revenue—a share that has grown from negligible levels two years ago, per Benzinga’s Q3 call transcript. The Semiconductor Solutions segment surged 127% year over year to $20.8 billion. Infrastructure Software, anchored by the VMware portfolio, grew 29% to $8.75 billion, slightly missing the $8.82 billion consensus estimate. However, it did not attract the concentrated analyst scrutiny that a software miss typically draws, unlike in the second quarter.

For the fourth quarter of fiscal 2026, Broadcom guided total revenue to approximately $34.8 billion, 93% above the same period last year. AI semiconductor revenue is projected to reach $21.7 billion—a 236% year-over-year increase—as XPU and networking revenue together triple year on year, per the earnings press release. The guidance landed roughly $230 million below the analyst consensus of $35.03 billion, enough to push AVGO down as much as 6% in after-hours trading before partially recovering. The stock closed the regular session Wednesday at $367.47, up approximately 6% year to date—significantly lagging the S&P 500’s 12% gain over the same period, a divergence that has puzzled observers given the company’s underlying AI chip growth trajectory.

Cody Acree, an equity research analyst at StoneX with a Buy rating on the stock, captured the market reaction precisely: “The magnitude is not quite enough from a top and bottom line standpoint on the beat and raise when you have a company that is this levered to AI,” he told Yahoo Finance. The board declared a quarterly cash dividend of $0.65 per share, payable September 30 to shareholders of record as of September 21.

**The Roadmap: $58B This Year, $115B Next, $230B in Two**

The fiscal 2027 and 2028 financial figures Tan disclosed Wednesday were prepared remarks rather than responses to analyst questions—a signal that Broadcom considers them sufficiently reliable to state publicly instead of letting them emerge from Q&A hedging.

Full-year fiscal 2026 AI semiconductor revenue is now projected at $58 billion, up from the prior $56 billion guidance—a 186% year-over-year increase, per Benzinga’s earnings call transcript. For fiscal 2027, Broadcom has secured the supply necessary to double AI revenue to approximately $115 billion. For fiscal 2028, the company has visibility to double again to $230 billion, with supply also secured for that target.

The crucial qualifier, stated plainly by Tan: demand for fiscal 2027 already exceeds the $115 billion outlook, per the Investing.com Q3 earnings transcript. Broadcom is not withholding a bullish forecast to appear conservative. It is capping the number because it has not yet secured all the supply chain capacity required to fulfill it. The forward demand roadmap across Broadcom’s six XPU customers extends to 30 gigawatts of aggregate compute capacity—a figure that maps to revenue well above the disclosed targets, per TheStreet’s live earnings call updates.

Whether Broadcom can convert that order pipeline into delivered silicon—and whether TSMC can expand fabrication capacity quickly enough—will define the AI infrastructure story for the next two years.

**What Actually Limits the $115B Target: CoWoS, HBM, and Power**

Tan identified the specific supply chain inputs pacing Broadcom’s AI revenue growth: leading-edge wafers at the 3-nanometer node, CoWoS advanced packaging substrates, high-bandwidth memory, power, land, and system components required for large-scale data center deployment, per Yahoo Finance’s Q3 earnings call highlights.

Understanding why packaging, rather than silicon, is the binding constraint requires a brief look at how modern AI chips are manufactured. A custom accelerator like Google’s Ironwood TPU is not merely a logic die; it is a system assembled through CoWoS (Chip-on-Wafer-on-Substrate), the advanced 2.5D packaging process TSMC developed over 12 years and now controls at roughly 90% of global capacity at AI-chip scale, according to Epoch AI’s chip supply research. CoWoS mounts the logic die on a silicon interposer alongside stacks of high-bandwidth memory, connecting them with thousands of short, dense electrical paths that enable the terabytes-per-second memory bandwidth modern AI model weights require. Without this packaging step, a fabricated 3nm wafer cannot become a functional, shippable accelerator.

TSMC has confirmed that CoWoS capacity remains sold out through 2026 and into 2027, with packaging lead times stretching 52 to 78 weeks. Scaling from approximately 35,000 CoWoS units per month in late 2024 to 130,000 by the end of 2026 represents an extraordinary manufacturing achievement—and it still falls short of demand. HBM memory from SK Hynix, Samsung, and Micron presents a parallel constraint: AI chip designers consumed roughly 90% of global HBM supply in 2025, leaving virtually no margin for demand growth without substantial new capacity additions.