AVGO · Forward model · Bull case
The Bull case, 20 quarters out
Each segment is projected from its own operating driver, rolled up into consolidated cash flow, and discounted back to a fair value per share. The assumptions below are editable — change them and every number on this page moves with them.
Quarters are labelled by the calendar year the fiscal quarter ENDS in, matching the rest of r40: 2026 Q2 is Broadcom's second fiscal quarter ended 3 May 2026. Broadcom reports two segments and this model carries three verticals, because the segment that matters splits into two businesses with almost nothing in common. Semiconductor Solutions and Infrastructure Software are reported on the face of every release ($15,009M and $7,178M in the basis quarter, $22,187M total). AI semiconductor revenue is disclosed separately in the CEO's quote every quarter but is NOT a reportable segment: $5.2B in 2025 Q3, $8.4B in 2026 Q1, $10.8B in 2026 Q2 are stated in dollars; 2025 Q2 and 2025 Q4 are given only as growth rates (+46% and +74% year over year) and are marked ESTIMATED here, derived from the +143% and the FY2024 base. Non-AI semiconductor is the arithmetic remainder of Semiconductor Solutions less AI, exact wherever AI is exact, so it inherits the two estimated flags and nothing more. Every quarter's three verticals sum to the reported consolidated total to the dollar. Nothing is apportioned. What is NOT split out: AI networking is roughly 40% of AI revenue on management's own description, but Broadcom has never published a quarterly series for it, so this model does not invent one — XPUs and AI networking ride in a single vertical and the split lives in the notes. What is assumed rather than reported: the segment margin and capex splits. Broadcom discloses consolidated Adjusted EBITDA ($15,244M, 68.7% of revenue) and consolidated capex ($231M) but not by segment, so the three margin inputs here are the author's and are calibrated to reproduce the consolidated number. Note also that the site's own revenue series carries $22,190M for the basis quarter against the press release's $22,187M; the actuals here follow the release.
This case is an argument about the asymptote, not about demand. It says the 17% quarterly decay in build rate is wrong — that inference at scale keeps pulling custom silicon in, that the customer list widens past the current six to sovereigns and the second tier of labs, and that co-packaged optics keeps AI networking's share of the fabric rising rather than flat. Note what it does NOT reach: even here the model does not get AI revenue to the $200B-a-year run rate the most aggressive sell-side numbers imply by 2030, because a compounding tilt cannot rebuild the capacity ramp underneath it. That gap is the finding, not a tuning target.
Latest: $105.31B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q2 | $15.00B |
| 2025Q3 | $15.95B |
| 2025Q4 | $18.02B |
| 2026Q1 | $19.31B |
| 2026Q2 | $22.19B |
| 2026Q3E | $29.69B |
| 2026Q4E | $36.38B |
| 2027Q1E | $42.37B |
| 2027Q2E | $47.79B |
| 2027Q3E | $52.74B |
| 2027Q4E | $57.31B |
| 2028Q1E | $61.56B |
| 2028Q2E | $65.54B |
| 2028Q3E | $69.31B |
| 2028Q4E | $72.91B |
| 2029Q1E | $76.37B |
| 2029Q2E | $79.72B |
| 2029Q3E | $82.99B |
| 2029Q4E | $86.20B |
| 2030Q1E | $89.37B |
| 2030Q2E | $92.53B |
| 2030Q3E | $95.69B |
| 2030Q4E | $98.86B |
| 2031Q1E | $102.06B |
| 2031Q2E | $105.31B |
What drives each segment
AI Semiconductors
Capacity × utilisation × priceCustom accelerators designed with a handful of hyperscalers and frontier labs, plus the Ethernet fabric that wires them together. Half the company already, and the only line whose slope decides whether the stock is cheap or expensive.
Latest: $68.12B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q2 | $4.44B |
| 2025Q3 | $5.20B |
| 2025Q4 | $6.44B |
| 2026Q1 | $8.40B |
| 2026Q2 | $10.80B |
| 2026Q3E | $16.60B |
| 2026Q4E | $21.88B |
| 2027Q1E | $26.65B |
| 2027Q2E | $30.97B |
| 2027Q3E | $34.88B |
| 2027Q4E | $38.42B |
| 2028Q1E | $41.65B |
| 2028Q2E | $44.60B |
| 2028Q3E | $47.30B |
| 2028Q4E | $49.79B |
| 2029Q1E | $52.11B |
| 2029Q2E | $54.26B |
| 2029Q3E | $56.28B |
| 2029Q4E | $58.19B |
| 2030Q1E | $60.01B |
| 2030Q2E | $61.74B |
| 2030Q3E | $63.41B |
| 2030Q4E | $65.02B |
| 2031Q1E | $66.59B |
| 2031Q2E | $68.12B |
Assumptions & reasoning
- Modelled as megawatts of AI compute Broadcom's silicon ships into each quarter, times the silicon content per megawatt, rather than as a growth rate. That is deliberate: the customer commitments are quoted in gigawatts, the constraint is advanced packaging and HBM allocation, and a single growth number would hide both. The megawatt count is BACKED OUT of revenue at $18M of content per megawatt, not disclosed — Broadcom publishes no units, no wafer starts and no gigawatts shipped.
- The calibration is checkable against management's own numbers. The first two projected quarters land at $16.0B and $20.4B against a guided $16.0B for 2026 Q3 and the $20.8B implied by a $56B fiscal 2026, and the four quarters of fiscal 2027 total $110B against a stated 'more than $100 billion'. The model clears the guide rather than matching it; that gap is the model being slightly generous in the guided window, not a forecast of a beat.
- Content per megawatt FALLS toward 1.75% a quarter once the drift engages, from $18.0M to $14.3M by 2031. This is the assumption doing the most work and the one most likely to be wrong in either direction. Custom silicon exists because it is cheaper per unit of work than a merchant GPU, and that logic does not stop applying to Broadcom's own pricing once a second ASIC vendor is in the room. If you think Broadcom holds content per megawatt flat, that slider adds roughly a third to the terminal AI line.
- Build rate decays 17% a quarter, so megawatts energised converge on about 2,560 rather than compounding forever. This is where the model refuses to follow the trend line: at the fiscal 2027 exit rate held flat, AI alone would pass $120B a year, and a model with no asymptote arrives at a number larger than the whole accelerator market. The decay is the discipline, and the Bull case is the argument that it is wrong.
- Concentration is the risk that no slider on this page expresses. Roughly six customers carry this line — Google, Meta, Anthropic, OpenAI, Apple and one more — and any single programme moving in-house, slipping a node or being cancelled takes a visible bite out of a business now worth half of Broadcom. Bookings above $30B in the basis quarter against $10.8B shipped is the offsetting fact: the backlog is real and it is contracted.
- Margin starts at 74% and glides to 68%. AI silicon is a high-margin business today because the design win is sticky and the customer pays for the engineering, but the terminal number assumes competition and customer bargaining power eventually cost Broadcom six points.
Non-AI Semiconductors
Growth pathBroadband, enterprise storage, non-AI enterprise networking, wireless and industrial: the businesses Broadcom was built on. Barely a fifth of revenue now, coming off a long downcycle, and the part of the company that behaves like a semiconductor company.
Latest: $11.95B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q2 | $3.96B |
| 2025Q3 | $3.97B |
| 2025Q4 | $4.63B |
| 2026Q1 | $4.12B |
| 2026Q2 | $4.21B |
| 2026Q3E | $4.59B |
| 2026Q4E | $4.96B |
| 2027Q1E | $5.31B |
| 2027Q2E | $5.66B |
| 2027Q3E | $6.00B |
| 2027Q4E | $6.34B |
| 2028Q1E | $6.68B |
| 2028Q2E | $7.03B |
| 2028Q3E | $7.37B |
| 2028Q4E | $7.73B |
| 2029Q1E | $8.09B |
| 2029Q2E | $8.46B |
| 2029Q3E | $8.85B |
| 2029Q4E | $9.24B |
| 2030Q1E | $9.65B |
| 2030Q2E | $10.08B |
| 2030Q3E | $10.52B |
| 2030Q4E | $10.98B |
| 2031Q1E | $11.46B |
| 2031Q2E | $11.95B |
Assumptions & reasoning
- Carried on a plain growth rate, and that is the honest choice here rather than a fallback. Five end-markets on five different cycles roll up into something that tracks the semiconductor cycle and nothing else; there is no single operating unit underneath worth pretending to model, and Broadcom discloses none.
- This line is a remainder, not a reported number: Semiconductor Solutions less the AI revenue stated in the CEO's quote. It is exact in the three quarters where the AI figure is given in dollars and derived in the two where only a growth rate was given, which is why 2025 Q2 and 2025 Q4 carry an estimated flag.
- 6.9% in the first projected quarter is the sequential step to the roughly $4.5B management pointed at for 2026 Q3 — a cyclical recovery, with the fiscal third quarter also being when the wireless content ramp begins. Bookings above $6B against $4.2B shipped in the basis quarter is the evidence that the recovery is real rather than seasonal.
- Terminal growth of 1.25% a quarter, about 5% a year, is deliberately unexciting. This is a mature analogue and connectivity portfolio with high share in markets that grow with units shipped; the recovery is a return to trend, not a new one.
- Margin is held flat at 55%, well below AI and well below software. Wireless in particular carries real bill-of-materials cost and a customer with unusual bargaining power, and there is no glide here because there is no story for why the mix improves.
- The understated risk: the 2025 Q4 spike to $4.6B is the seasonal handset ramp, and it is the same customer that now buys custom silicon from the AI line. Concentration in this model is not confined to the vertical that advertises it.
Infrastructure Software
Growth pathVMware, essentially. A subscription estate being converted from perpetual licences to VMware Cloud Foundation at much higher prices, bolted onto the mainframe and security assets from CA and Symantec. Boring, and the reason Broadcom's cash flow does not swing with the silicon cycle.
Latest: $25.23B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q2 | $6.60B |
| 2025Q3 | $6.79B |
| 2025Q4 | $6.94B |
| 2026Q1 | $6.80B |
| 2026Q2 | $7.18B |
| 2026Q3E | $8.49B |
| 2026Q4E | $9.54B |
| 2027Q1E | $10.41B |
| 2027Q2E | $11.16B |
| 2027Q3E | $11.86B |
| 2027Q4E | $12.54B |
| 2028Q1E | $13.22B |
| 2028Q2E | $13.92B |
| 2028Q3E | $14.64B |
| 2028Q4E | $15.39B |
| 2029Q1E | $16.17B |
| 2029Q2E | $17.00B |
| 2029Q3E | $17.86B |
| 2029Q4E | $18.76B |
| 2030Q1E | $19.71B |
| 2030Q2E | $20.71B |
| 2030Q3E | $21.76B |
| 2030Q4E | $22.86B |
| 2031Q1E | $24.02B |
| 2031Q2E | $25.23B |
Assumptions & reasoning
- Every actual here is a reported segment figure, taken straight off the face of the earnings release. Nothing in this vertical's revenue history is estimated or apportioned — what is assumed is the cost side, since Broadcom does not publish segment EBITDA or segment capex.
- The model does NOT chase the guided quarter, and this is the one place it deliberately falls short. Total 2026 Q3 revenue is guided to roughly $29.4B and AI to $16.0B; with non-AI semiconductors near $4.5B the residual leaves about $8.9B for software, a 24% sequential jump for a business that grew 1% year over year two quarters ago. This line projects 16% instead, landing near $8.3B, and consolidated revenue therefore comes in about 2% under the guide. Treating part of that step as renewal timing rather than run-rate is a judgement call, and it is stated rather than tuned away.
- The whole growth story is renewal repricing, not new customers. Perpetual VMware licences convert to VCF subscriptions at a large multiple of the old price as each contract comes up, so growth is a function of how much of the estate is still unconverted. That pool is finite, which is why the rate decays to 2% a quarter — roughly 8% a year — rather than holding.
- 82% EBITDA margin, gliding to 80%, is the highest on this page and it is what makes the mix argument work: as AI margin drifts down, software is the ballast. Broadcom does not disclose segment profitability, so this number is calibrated so the three verticals less 4% central overhead reproduce the consolidated 68.7% Adjusted EBITDA margin actually reported in the basis quarter.
- The risk here is churn, and it is not visible in a quarterly revenue line until it is large. Repricing an estate at a multiple of the old cost is a strong incentive for enterprises to migrate to a hyperscaler or to a competing hypervisor; departures show up years later as a renewal that does not happen. The Bear case is where that lives.
Where each case comes from
Hock case — primary sources
The primary sources this case is built from — filings, calls and posts. Where they are claims by an interested party rather than disclosures, every number in the Hock column is what happens if they are taken at face value.
What management has actually put a number on
- Jun 3, 2026 Q2 semiconductor revenue from AI of $10.8 billion grew 143% year-over-year, above our forecast, driven by increasing demand for custom AI accelerators and AI networking.
- Jun 3, 2026 The momentum continues and in Q3 we expect semiconductor revenue from AI to grow over 200 percent year-over-year to $16.0 billion.
- Jun 3, 2026 In Q3 we expect consolidated revenue growth to increase 84% year-over-year to $29.4 billion, with non-GAAP operating margin stable at 67% reflecting our strong operating leverage.
The ramp behind it, quarter by quarter
- Dec 11, 2025 We see the momentum continuing in Q1 and expect AI semiconductor revenue to double year-over-year to $8.2 billion, driven by custom AI accelerators and Ethernet AI switches.
- Mar 4, 2026 Our AI revenue growth is accelerating, and we expect AI semiconductor revenue to be $10.7 billion in Q2.
- Sep 4, 2025 We expect growth in AI semiconductor revenue to accelerate to $6.2 billion in Q4, delivering eleven consecutive quarters of growth, as our customers continue to strongly invest.
Who the demand is coming from
- Jun 3, 2026 Tan and Brockman unveil their first jointly designed custom chip and address whether AI capex is a bubble, arguing committed customer demand underwrites the build-out.
- Jun 5, 2026 Tan describes the early Anthropic custom-chip engagement as a leap of faith that turned into a great bet, and discusses AI scaling and revenue visibility.
From cash flow to fair value
| Present value of free cash flow, 20 quarters | $662.12B |
| Terminal-year revenue | $401.91B |
| Terminal-year EBITDA | $277.23B |
| Exit multiple, on revenue | 11.0x |
| Terminal value | $4.42T |
| Discounted at 8.5% a year, terminal value becomes | $2.94T |
| Enterprise value | $3.60T |
| Net cash | -$45.28B |
| Equity value | $3.56T |
| Diluted shares | 4.88B |
| Fair value per share | $729.50 |
| Against the current price of $364.03 | +100% |
7x terminal revenue against roughly 19x trailing enterprise value to revenue today. The terminal year is the awkward part, exactly as it should be: by 2031 this model has consolidated revenue flat at about $222B a year at a 66% EBITDA margin, so 7x revenue is about 10.6x EBITDA for a business that has stopped compounding. That is fair-to-generous for a plateau, and it is a deliberate choice not to argue the plateau away inside the multiple — the Bull case argues about the ramp instead, which is where the disagreement actually lives. Move the slider and watch how little else matters: the base case is $195 at 4x, $282 at 7x and $369 at 10x, while moving the discount rate from 8% to 13% only takes the answer from $300 to $246. One input swings this page more than every operating assumption on it combined, and at today's price the market is paying about 10x that flat terminal year. What the model does NOT count: the roughly $65B of gross debt is netted off as a $45B net cash deficit, but the interest on it — about $700M a quarter in cash — is not modelled as a separate charge, so free cash flow here is pre-interest and the equity value is correspondingly flattered by a few percent. Nor does it count the buyback: Broadcom announced a new $10B repurchase programme in the basis quarter, and share count is held constant at the 4,876M diluted shares reported.
Read the other way round: at $364.03 the market is paying 4.3x terminal-year revenue, holding every other assumption on this page fixed. That is the number to argue about.
The projected path
| Quarter | AI Semiconductors | Non-AI Semiconductors | Infrastructure Software | Revenue | YoY | EBITDA | Capex | FCF | R40 | PV of FCF |
|---|---|---|---|---|---|---|---|---|---|---|
| 2026 Q3E | $16.60B | $4.59B | $8.49B | $29.69B | +86% | $21.36B | $381M | $18.04B | +147 | $17.68B |
| 2026 Q4E | $21.88B | $4.96B | $9.54B | $36.38B | +102% | $26.09B | $501M | $22.01B | +162 | $21.13B |
| 2027 Q1E | $26.65B | $5.31B | $10.41B | $42.37B | +119% | $30.27B | $618M | $25.50B | +180 | $23.98B |
| 2027 Q2E | $30.97B | $5.66B | $11.16B | $47.79B | +115% | $33.99B | $731M | $28.60B | +175 | $26.36B |
| 2027 Q3E | $34.88B | $6.00B | $11.86B | $52.74B | +78% | $37.36B | $840M | $31.40B | +137 | $28.36B |
| 2027 Q4E | $38.42B | $6.34B | $12.54B | $57.31B | +58% | $40.44B | $944M | $33.96B | +117 | $30.05B |
| 2028 Q1E | $41.65B | $6.68B | $13.22B | $61.56B | +45% | $43.28B | $1.04B | $36.33B | +104 | $31.49B |
| 2028 Q2E | $44.60B | $7.03B | $13.92B | $65.54B | +37% | $45.94B | $1.14B | $38.53B | +96 | $32.73B |
| 2028 Q3E | $47.30B | $7.37B | $14.64B | $69.31B | +31% | $48.45B | $1.23B | $40.61B | +90 | $33.80B |
| 2028 Q4E | $49.79B | $7.73B | $15.39B | $72.91B | +27% | $50.84B | $1.31B | $42.59B | +86 | $34.74B |
| 2029 Q1E | $52.11B | $8.09B | $16.17B | $76.37B | +24% | $53.14B | $1.40B | $44.50B | +82 | $35.56B |
| 2029 Q2E | $54.26B | $8.46B | $17.00B | $79.72B | +22% | $55.37B | $1.48B | $46.35B | +80 | $36.28B |
| 2029 Q3E | $56.28B | $8.85B | $17.86B | $82.99B | +20% | $57.54B | $1.55B | $48.15B | +78 | $36.94B |
| 2029 Q4E | $58.19B | $9.24B | $18.76B | $86.20B | +18% | $59.69B | $1.63B | $49.93B | +76 | $37.53B |
| 2030 Q1E | $60.01B | $9.65B | $19.71B | $89.37B | +17% | $61.81B | $1.70B | $51.70B | +75 | $38.07B |
| 2030 Q2E | $61.74B | $10.08B | $20.71B | $92.53B | +16% | $63.93B | $1.77B | $53.46B | +74 | $38.58B |
| 2030 Q3E | $63.41B | $10.52B | $21.76B | $95.69B | +15% | $66.06B | $1.84B | $55.23B | +73 | $39.05B |
| 2030 Q4E | $65.02B | $10.98B | $22.86B | $98.86B | +15% | $68.21B | $1.90B | $57.02B | +72 | $39.50B |
| 2031 Q1E | $66.59B | $11.46B | $24.02B | $102.06B | +14% | $70.38B | $1.97B | $58.83B | +72 | $39.93B |
| 2031 Q2E | $68.12B | $11.95B | $25.23B | $105.31B | +14% | $72.59B | $2.03B | $60.68B | +71 | $40.35B |
Every row is projected. A year-over-year change is shown only where the quarter it compares against exists — an em dash means there is no comparable quarter, not a flat year.
Model revisions
Assumptions are marked to reality as each quarter prints. Every change is appended here, with the fair value the model produced at the time, so the model's own history stays visible.
| Date | Changed | Fair value then | Note |
|---|---|---|---|
| 2026-08-20 | all | $282.09 | First cut, built on the 2026 Q2 earnings release. Three verticals: AI semiconductors on a megawatts-shipped driver calibrated to the $56B fiscal 2026 and $100B-plus fiscal 2027 AI targets, non-AI semiconductors on a cyclical recovery, and Infrastructure Software on a decaying VMware repricing uplift. |