Skip to content
FrontierPicks

Watchlist

AVGO · Broadcom Inc.

Conviction · LOW Cyclical recovery Catalyst · AI chips & memory

Last analysed ·

Against its published line

Nothing is through its line on this close.

How to read this

The red mark is the published kill line — the price that would prove the pick wrong. The dot is where the name closed on 10 September 2026; a dot LEFT of the mark has closed through its line.

Distance is drawn on a square-root scale, so close calls get the room. Past 8% a row stops competing and reads well clear, with a hollow dot to say the figure is off the drawn scale. Rows run tightest first.

How a pick resolves

AVGOBroadcom Inc.
$340.00
$360.83
+6.1%

Current thesis

Broadcom beat and raised on 2026-09-02 — Q3 revenue $29.591B vs $29.362B consensus, AI semis $16.7B (+221%), FY26 AI lifted to $58B, plus a stated ~$115B FY27 / ~$230B FY28 AI roadmap — but Q4 revenue guided $34.8B against $35.03B consensus, Q4 gross margin to ~73% from 78%, and the shares fell about 5% after hours from a $367.24 close with RSI(14) at 24.1.

Kill line

A weekly close below $340 completes the failed $370 shelf and takes out the 2026-08-26 close of $355.59. Secondary: Q4 AI semiconductor revenue landing under the guided $21.7B at the December print, or the ~$115B FY27 line-of-sight figure not being reaffirmed.

Pick status

Open commitment catalyst 3d agoscored if the kill line above fires How this is scored →

Latest analysis and events for AVGO —

As of 3 September 2026, the latest FrontierPicks analysis for Broadcom Inc. (AVGO): 26 July 2026: Samsung–Broadcom ~$200B AI-chip pact through 2030 — second-source foundry away from TSMC, extends custom-silicon roster to Google/Meta/OpenAI/Apple/Samsung. Watch volume-ramp and node-allocation follow-through vs the headline.

Kill line: A weekly close below $340 completes the failed $370 shelf and takes out the 2026-08-26 close of $355.59. Secondary: Q4 AI semiconductor revenue landing under the guided $21.7B at the December print, or the ~$115B FY27 line-of-sight figure not being reaffirmed.

Most recent dated event on file: — catalyst 3d ago.

Refresh note — 2026-09-03. Advances the 2026-09-02 pre-print read through the print itself. Last completed daily close in the graded series: $367.24 (2026-09-02, before the after-close release), RSI(14) 24.1, 23.6% below the $480.81 52-week high.

AVGO — Broadcom Inc.

Current Thesis

The leg being bought is unchanged since coverage opened in April: Broadcom designs the custom accelerators for the largest AI buyers outside NVIDIA's merchant line — Google, Meta, OpenAI, Anthropic — sells the Tomahawk/Jericho Ethernet fabric between them, and carries the VMware software annuity underneath. On 2026-09-02 that thesis got its numbers. Q3 FY26 revenue landed at $29.591B against a $29.362B consensus, adjusted EPS $3.32 against $3.24, and AI semiconductor revenue $16.7B (+221% YoY) against the $16.0B guided on 2026-06-03. XPUs made up 73% of the AI line with shipments up more than 3.5x YoY; AI networking rose more than 2.5x. FY26 AI revenue was raised to $58B from $56B.

The same call produced the first quantified multi-year figures management has put on the record: roughly $115B of AI semiconductor revenue in FY27 and roughly $230B in FY28, described by Hock Tan as "a judged number" summing to about $350 billion across the two years, with content per gigawatt of $20–30B across six XPU customers, and a stated target to exceed $30 of EPS in FY2028.

The tape took the other side. Q4 revenue was guided to $34.8B (+93% YoY) against the $35.03B LSEG consensus, Q4 gross margin to around 73% versus 78% a year earlier, and CNBC reported the shares down about 5% in extended trade. So the narrative is saturated — the largest upward revision of this cycle printed on 2026-09-02 into daily mainstream preview coverage that had run from 2026-08-29, and the immediate bid widened for nobody; attention is fully present and price is 23.6% under its high.

Bullish and bearish views on Broadcom Inc.

The model's bull view on Broadcom Inc. (AVGO), in brief: 2026-09-02 — the guide was beaten, not met. The bear view: 2026-09-02 — the first guide of this cycle to land under the Street. Both cases follow in full.

Bull Case

  • 2026-09-02 — the guide was beaten, not met. AI semiconductor revenue came in $0.7B above the $16.0B set on 2026-06-03, and total revenue cleared the $29.362B Street compilation.
  • 2026-09-02 — FY26 AI revenue raised to $58B from $56B (+186% YoY), the first revision to a number fixed for thirteen weeks.
  • 2026-09-02 — Q4 AI semiconductors guided to $21.7B (+236% YoY), a sequential step up from the $16.7B just reported, with infrastructure software at approximately $8.7B and operating margin around 66%.
  • 2026-09-02 — the second-sourcing objection was answered with named programs. Production shipments of Google's TPU v8i began during Q3 ahead of schedule with high-volume shipments guided for Q4; three generations of Meta's MTIA accelerator are scheduled through end-2027; Anthropic is stated to become the largest XPU customer in 2027 and remain so in 2028; OpenAI's "Jalapeno" was described as on plan.
  • 2026-09-02 — the multi-year frame is now numeric. Roughly $115B FY27 and $230B FY28 AI revenue, against sell-side price targets set before the print at $580 (JPMorgan, 2026-08-31), $550 (Bernstein), $530 (BofA) and a $525.97 average across 49 analysts polled by S&P Global.
  • 2026-08-26 — the funding pool was measured a week earlier. NVIDIA reported Q2 FY27 revenue $96.2B, Data Center $89.0B, 75.0% gross margin, and guided Q3 to $108.0B ±2% with no China Data Center compute revenue assumed.

Bear Case

  • 2026-09-02 — the first guide of this cycle to land under the Street. $34.8B against $35.03B (LSEG). The shortfall is small in dollars and decisive in framing for a name that had cleared every prior bar.
  • 2026-09-02 — gross margin steps to roughly 73% in Q4 from 78% a year ago, with management defending operating margin rather than gross margin as XPU mix dilutes it, and Q4 capex of $1.4B for capacity.
  • 2026-09-02 — the software annuity flattens. Infrastructure software was described as stabilising at approximately $8.7B in Q4, against the $8.9B guided for Q3 on 2026-06-03.
  • 2026-09-02 — the FY28 $230B figure is management's judgement of forward shipment, not converted backlog, and it rests on six customers, one of which (Anthropic) is a private company scheduled to become the single largest.
  • 2026-09-01 — the merchant incumbent moved on the adjacency. NVIDIA invested $3.5B with MediaTek to extend NVLink Fusion into custom AI silicon, and a Counterpoint analyst framed NVHBM as a move into Broadcom's territory (Benzinga, 2026-09-01).
  • 2026-06-03 and 2026-09-02 — beats keep getting sold. Benzinga dated a 12.6% decline to the prior beat; this one drew a reported ~5% extended-session decline.

Setup & Price Structure

Price went into the print already broken relative to the structure this coverage has tracked. The $370 July shelf and the $369.20 area Benzinga marked as the 200-day SMA on 2026-08-24 both sat above the 2026-09-02 close of $367.24. RSI(14) at 24.1 is the lowest reading of the covered period, below the 26.1 recorded on 2026-09-01; the three-month price change is -12.2% and the shares are 23.6% under the $480.81 high.

Applying CNBC's reported ~5% extended-session decline to that close places the first post-print quote in the high-$340s to low-$350s — an inference from a reported percentage, not a printed close, and the 2026-09-03 session had not completed at writing. The nearest reference below is the 2026-08-26 close of $355.59; the published break level of $340 sits under that.

Positioning observables, stated as observables: options priced a 7.3% post-earnings move (Investing.com, 2026-08-26, citing Bloomberg-compiled data), the actual reported reaction was smaller; mainstream preview coverage ran daily from 2026-08-29 through 2026-09-01 on a guide that had not changed since 2026-06-03; price trades below a 200-day marker it lost in late August; sell-side targets cluster $525–580 against a quote roughly 35% lower, a dispersion that has widened rather than closed since 2026-08-20; insider sales run through Rule 10b5-1 plans (Samueli's adopted 2025-12-16), so their timing carries no read on price. Nothing in this structure is a base — the shelf that had held four sessions in late August is now above price rather than under it.

Catalyst Calendar (next 30 days)

  • ~2026-12-11 (est.) — Q4 FY26 results. Beyond the 30-day window and worth stating plainly: after 2026-09-02 there is no company-dated event for roughly thirteen weeks.

Elapsed catalysts

  • ~2026-09-08 (est.) — Q3 FY26 Form 10-Q. Updates the maximum theoretical loss disclosure on the AI XPV residual-value guarantees and shows whether they extend to the newly reported senior tranche. (passed 3d ago)
  • ~2026-09-10 (est.) — TSMC August monthly revenue. July printed NT$467.58B on 2026-08-10, +44.7% YoY versus June's +67.9%; August tests whether that step-down was a base effect. (passed 1d ago)
  • ~2026-09 (no announced date) — Pricing or allocation on the SPV AI-chip debt financing. The window implied around 2026-08-31 passed without a print. (passed 11d ago)
  • ~2026-09 (no announced date) — Commerce decision on the expanded semiconductor tariff scope reported 2026-08-27, specifically whether data-centre servers keep the exemption attached to January 2026's 25% advanced-computing-chip duty. (passed 15d ago)

What Would Change Our Mind

The binary this coverage was pointed at has resolved, and it resolved with the fundamental number up and the price down. What breaks the thesis from here is the structure failing to hold the post-print zone: a weekly close below $340 takes out the 2026-08-26 close of $355.59 and completes the failure of the $370 shelf, leaving no marked level between there and the spring range. On the fundamental side, Q4 AI semiconductor revenue landing under the guided $21.7B at the December print, or management declining to reaffirm the ~$115B FY27 figure, would remove the only new information this print added. A restatement of the $115B/$230B numbers as bookings rather than shipped revenue would do the same.

The opposite reading is equally gradeable: a weekly close back above $370 reclaims the shelf and the 200-day area lost in late August, and would mark the point where the raised FY26 AI number began to set price again.

Correlation Notes

AVGO trades against the same hyperscaler capex line as NVDA (Q2 FY27, 2026-08-26: revenue $96.2B, Data Center $89.0B, Q3 guide $108.0B ±2%) and MRVL (Q2 FY27 revenue $2.739B, Q3 guide $3.15B ±5%; the shares fell nearly 4% on 2026-09-01 on a softer margin outlook). TSMC's monthly revenue is the cleanest upstream loading proxy. Two non-fundamental correlations have been active: private-credit appetite for AI infrastructure paper, via the SPV whose senior tranche Broadcom would partially guarantee, and US policy headlines — the 2026-08-28 session fell on tariff news with no company datapoint, and data-centre politics produced separate headlines on 2026-08-31 and 2026-09-01.

Notes

  • 2026-04-19: AI networking (Tomahawk/Jericho) + VMware cashflow — #2 behind NVDA
  • 2026-07-26: Samsung–Broadcom ~$200B AI-chip pact through 2030 — second-source foundry away from TSMC, extends custom-silicon roster to Google/Meta/OpenAI/Apple/Samsung. Watch volume-ramp and node-allocation follow-through vs the headline.
  • Broadcom reports a month behind the calendar-quarter cohort: Q3 FY26 landed 2026-09-02; Q4 FY26 falls in December, leaving ~13 weeks with no company-dated event.
  • Two businesses in one line. Infrastructure software is guided to roughly $8.7B in Q4 FY26; a semis-only read misses a material share of revenue.
  • The stated FY27 ~$115B and FY28 ~$230B AI figures are management's forward judgement of shipments, not contracted backlog converted to revenue.
  • AI XPV residual-value guarantees are contingent liabilities disclosed in the 10-Q; the senior debt sits in a special-purpose vehicle outside the balance sheet.
  • Customer concentration is structural — six XPU customers, with Anthropic stated to become the largest in 2027 — so one socket decision moves the segment more than macro.
  • Most insider sales run through Rule 10b5-1 plans (Samueli's adopted 2025-12-16), so their timing is pre-scheduled rather than a read on price.

Related · shared themes

See also · stocks to watch