Skip to content
Today
Sections

Invest Also reported elsewhere2 publishers 2 min read Published Updated

Hock Tan defends Broadcom's $230 billion fiscal 2028 AI forecast on inference demand

Dario Amodei's call for slower model development knocked 4.8% off Broadcom on Monday, and Hock Tan told CNBC his $115 billion and $230 billion AI chip targets for fiscal 2027 and 2028 have not moved. Anthropic is due to become his largest custom chip customer next year.

The Investor · Invest desk

Send a correction

What happened

  • Dario Amodei's weekend essay argued for moderating the pace of model development, drew support from Sam Altman and Elon Musk, and sent AI infrastructure investors back to their compute demand assumptions.
  • Broadcom shares fell 4.8% on Monday and the iShares Semiconductor ETF, a broad basket of chip stocks, fell 5.6%, with data-center component suppliers also hit.
  • Those forecasts, issued on the Sept. 2 fiscal third-quarter call, put AI semiconductor revenue at $115 billion in fiscal 2027 and $230 billion in fiscal 2028.

Why it matters

  • What it limits constraint Tan defended the targets on inference and said he does not know about training, so the executive who set the number has not publicly committed to the accelerator demand that Amodei's plan would actually slow.
  • Who's exposed exposure Broadcom's largest single custom chip line in 2027 and 2028 belongs to a lab whose chief executive is campaigning to slow the work those chips do.
  • What doesn't add up contradiction Tan endorses safeguards while insisting AI will not run wild by itself, so his disagreement with Amodei is about danger, and the compute forecast follows from that reading.
  • Who has to decide decision Having restated the figures twelve days after issuing them, Tan has left the next earnings call as the only place a cut can credibly appear.

Tan gave Jim Cramer two answers about the same forecast. "No, not in the least," he said when asked whether anything in the AI slowdown debate had caused him to reconsider Broadcom's fiscal 2027 and 2028 AI semiconductor forecasts [4]. On what underwrites those numbers he was narrower: "I don't know about training, but when you want to productize inference, I see it continuing to be very, very strong," Tan said [6]. Amodei's essay argues for moderating the pace of model development [7], and CNBC describes inference as the day-to-day usage of models after they have been trained [11].

Tan vouched for demand that shows up after training is finished.

The targets are $115 billion of AI semiconductor revenue in fiscal 2027 and $230 billion in fiscal 2028, both set out on the September 2 earnings call [1]. The second figure is twice the first, so it requires another $115 billion of AI revenue inside a single fiscal year [17]. Broadcom counts custom accelerators and networking chips in that line [2]. Tan said Anthropic is on track to be the largest custom chip customer in 2027 and to hold that position in 2028 [20], ahead of Google, whose tensor processing units Broadcom co-designs [9]. CNBC did not report Anthropic's share of the $115 billion.

Monday's selling was less discriminating than the story suggests. Broadcom fell 4.8% while the iShares Semiconductor ETF fell 5.6% [10], which means the company whose incoming largest custom customer wrote the essay dropped 0.8 percentage points less than the basket [18].

Tan said he agrees with Amodei about the need for some restrictions on AI [12]. "Like any tool, it's important to put governances, safeguards on how we use the tool," he told Cramer [13]. He also said, "It's not a live animal that will run wild by itself" [14], and likened generative AI to the Industrial Revolution that began in England in the 18th century [15].

If inference volume grows the way Tan describes, fiscal 2027 survives a slower frontier training cycle and the doubling into 2028 turns into a supply question. Should Anthropic act on its own three-step proposal and trim accelerator orders [8], the customer concentration Tan himself described is how that reaches the forecast. The third case is that the labs change nothing and Monday's 4.8% was noise [10]. I would weight the first, partly because Tan restated the target twelve days after issuing it [19]. A narrowed or withdrawn $115 billion at the next earnings call settles it the other way.

What to watch

  • Whether Broadcom restates the $115 billion fiscal 2027 figure at its next earnings call or narrows the range.
  • Any disclosure of committed custom accelerator volumes sitting behind the $230 billion fiscal 2028 number.
  • Whether Anthropic's own capital spending follows the three-step plan Amodei set out.

Information, not investment advice. This story is general information about companies and markets, drawn from cited reporting. It is not a recommendation to buy, sell or hold any security; check the cited sources before acting.

Clarity's read

What the record supports and how the coverage leans. The claims behind it follow.

Reality

Evidence62
Adoption55
Hype gap+18
Incentives82
Confidence64
Why these scores

Claim ledger

Ranked by verification strength, evidence, and original report placement.

  1. [1]

    On Broadcom's fiscal 2026 third-quarter earnings call on Sept. 2, Tan forecast AI semiconductor revenue of $115 billion in fiscal 2027, before doubling again to $230 billion in fiscal 2028.

    Reported SupportedView cited source
  2. [2]

    Broadcom's AI revenue consists of both custom AI accelerators and networking chips used in AI systems.

    Reported SupportedView cited source
  3. [3]

    Broadcom CEO Hock Tan told CNBC on Monday that the company stands by its long-term revenue targets, brushing off concerns about what a potential slowdown in frontier AI model development could mean for its business.

    Reported SupportedView cited source
  4. [4]

    "No, not in the least," Tan said on "Mad Money," when Jim Cramer asked whether anything in the AI slowdown debate has caused him to reconsider Broadcom's fiscal 2027 and 2028 AI semiconductor forecasts.

    Reported SupportedView cited source
  5. [5]

    Tan said: "We see the demand for compute infrastructure, for AI development or AI frontier models, and inference for the products that they feed to the world, as continuing to be very strong and, I believe, very durable."

    Reported SupportedView cited source
  6. [6]

    Tan said: "I don't know about training, but when you want to productize inference, I see it continuing to be very, very strong."

    Reported SupportedView cited source
  7. [7]

    Anthropic CEO Dario Amodei's weekend essay advocated for a moderation in the pace of model development, and his message received support from OpenAI CEO Sam Altman and Elon Musk, prompting investors in AI infrastructure providers to reconsider their expectations for compute demand during Monday's session.

    Reported SupportedView cited source
  8. [8]

    Amodei proposed a three-step plan to slow the pace of development without "sacrificing commercial advantage or the United States' lead in AI."

    Reported SupportedView cited source
  9. [9]

    Google has historically been considered Broadcom's biggest custom customer, co-designing Google's tensor processing units.

    Reported SupportedView cited source
  10. [10]

    Shares of Broadcom fell 4.8% on Monday, and the iShares Semiconductor ETF, a broad basket of chip stocks, fell 5.6%.

    Reported SupportedView cited source
  11. [11]

    CNBC describes inference as the day-to-day usage of AI models after they have been trained.

    Reported SupportedView cited source
  12. [12]

    Tan said he agrees with Amodei about the need for some restrictions on AI, but suggested he was less alarmed about where the technology is headed.

    Reported SupportedView cited source
  13. [13]

    "Like any tool, it's important to put governances, safeguards on how we use the tool," Tan told Cramer.

    Reported SupportedView cited source
  14. [14]

    Tan argued, "It's not a live animal that will run wild by itself."

    Reported SupportedView cited source
  15. [15]

    Tan said "AI, generative AI, the creation of those frontier models ... will create huge value," likening it to the Industrial Revolution that began in England in the 18th century.

    Reported SupportedView cited source
  16. [16]

    CNBC published the Tan interview report on Sept. 14, 2026.

    Reported SupportedView cited source
  17. [17]

    The fiscal 2028 target of $230 billion is twice the fiscal 2027 target of $115 billion, and requires an additional $115 billion of AI semiconductor revenue within one fiscal year.

    Derived Supported
  18. [18]

    Broadcom's 4.8% Monday decline was 0.8 percentage points smaller than the iShares Semiconductor ETF's 5.6% decline.

    Derived Supported
  19. [19]

    Tan reaffirmed the fiscal 2027 and 2028 targets twelve days after issuing them on the Sept. 2 earnings call.

    Derived Supported
  20. [20]

    Tan said Anthropic is on track to become Broadcom's largest custom chip customer in 2027 and to sustain that designation in 2028.

    Reported InsufficientView cited source
  21. [21]

    Broadcom's custom-chip business counts Anthropic as one of its most important customers.

    Reported InsufficientView cited source

Sources

2 independent publishers whose own reporting we read for this story.

  1. cnbc.com

    1 article · September 14, 2026

    Broadcom CEO addresses Anthropic's slowdown push, says AI revenue targets haven't changed
  2. seekingalpha.com

    1 article · September 15, 2026

    Broadcom's Biggest Customers Want To Slow AI Down - They're Still Buying

Topics and entities

Topics

Entities