The Best AI Buying Opportunity I've Seen in Months Is Hiding Here

Finance-focused summary (markets, thesis, valuation, risk)

Assets / tickers / instruments mentioned

• Broadcom (AVGO) (primary focus; discussed with market cap, P/E, moving averages, DCF price target)
• NVIDIA (NVDA)
• Meta Platforms (META)
• Alphabet / Google (GOOGL)
• AMD (AMD)
• MercadoLibre (MELI)
• Nu Holdings (NU)
• DiDi (referred to as “DiDi Local”)
• ByteDance
• Oracle (risk example)
• Arm
• OpenAI
• Anthropic
• CoreWeave
• Nebius / Neurius (spelled “Nebius/Neurius” in the subtitles)
• Anthropic chip/AI infrastructure financing structure involving banks (including Bank of America)
• Moving averages (200-day, 150-day, 50-day) referenced for AVGO technical context

Macro / market backdrop (short)

The video references a “start of the week” market tape:

• Meta +2%
• Broadcom ~+2%
• Brazil election-related optimism:
  • Nu +13%
  • DiDi Local +7.3%
  • MELI +8.2%
• Future catalyst timing:
  • Brazil final round of elections: Oct 25

---

Investment thesis for Broadcom (AVGO)

The presenter argues Broadcom is an “AI buying opportunity” due to attractive valuation versus expected growth and improving margins / free cash flow (FCF).

Valuation framing / key claims

• Broadcom market cap cited: ~$1.7 trillion
• Valuation concern addressed:
  • Historical P/E cited: ~46 (past year)
  • Forward P/E: ~20.8 (based on analyst expectations)

• Growth and per-share target logic:

  • Analyst/management target: $30 profit by 2028
  • Current stock price cited: ~$360
  • If 2028 profit is multiplied by 20, implied price: ~$600 per share (with the caveat that using a lower multiple could still exceed the current price)

• PEG ratio cited: ~0.5, used as evidence of undervaluation

Analyst expectations for AI revenue growth (explicit numbers)

The video contrasts market expectations with management’s AI roadmap:

Market expectations (for Broadcom’s FY):

• FY26: +65.8%
• FY27: +64%
• FY28: +56.3%

Management AI revenue guidance:

• AI revenue doubles in FY27 to $115B
• AI revenue doubles again in FY28 to $230B

Presenter interpretation

• $230B is described as AI revenue only, not total company revenue; total revenue would also include non-AI semiconductor + infrastructure software.

Revenue mix (company financial components)

Broadcom is split into:

• Semiconductor revenue: $20.8B last quarter
• Infrastructure software revenue: $8.75B (smaller than semis, but still meaningful)

Margin / operating quality angle (qualitative, but direction is clear)

The presenter claims “everything is moving up and to the right,” specifically:

• Operating margin
• Free cash flow margin
• Gross margin
• Net margin
• EBITDA margin

Conclusion in the model: no sign the business slows “anytime soon” (as argued by the presenter).

---

AI/semiconductor “financing” catalyst: Broadcom ↔ Anthropic (risk + upside)

The presenter highlights Broadcom’s role in an AI chip lease/financing structure for Anthropic:

Broadcom will act as a guarantor for a financing arrangement where banks (including Bank of America) lend Anthropic up to $42B to lease chips. If Anthropic can’t repay, Broadcom is expected to “put up the money.”

Possible second-order upside

• Broadcom “may also acquire a stake” in Anthropic later
• Framed as “cyclical financing” similar to NVIDIA—i.e., scale-up funded via partners/capex cycles—contingent on Broadcom’s ability to maintain strong free cash flow and demand for its chips

Explicit risk: customer / concentration risk

• Risk highlighted: concentration if Anthropic/OpenAI become very large customers and growth expectations fail.
• The presenter warns this approach can be problematic “if one…can’t grow as fast as expected” and can’t meet commitments, using Oracle as an example.

---

Technical / positioning notes (AVGO chart context)

• Intraday: AVGO trading below 200-day and 150-day moving averages, but trending toward the 200-day
• Weekly chart: AVGO fell below and stayed below the 50-day moving average, needing to regain it for momentum confidence

---

DCF methodology & framework (as presented)

Scenario structure

• DCF uses pessimistic / base / optimistic cases
• Weighted outcome: ~$527
• Weighting:
  • 50% baseline
  • 30% optimistic
  • 20% pessimistic
• Presenter encourages using an assumptions tab and sensitivity tab to adjust inputs

Base-case DCF timeline and key drivers (explicit)

• “Year 1” in the model is FY2027 (because it’s near the end of FY2026)
• AI semiconductor revenue assumptions:
  • FY2027: $115B
  • FY2028: $230B
• Total revenue framework (with non-AI items):
  • Non-AI semiconductor + infrastructure software reaching $174.9B (noted as +60% increase)
• Revenue growth slows materially by FY2029 in the base scenario:
  • Growth rate ramp down sequence mentioned: ~22% to 14.9, 7.8, 6%, 5.1, down to 4.9
• Margin assumption in base narrative:
  • Presenter claims base path supports EBITDA margin up to ~57% (could be lower in the pessimistic case)

Optimistic vs pessimistic scenario logic (qualitative)

• Optimistic: AI semis revenue stays the same in the first two years (could be higher) → higher growth + higher margins
• Pessimistic: targets missed → lower growth + lower margins

Sensitivity analysis inputs (explicit options)

• The model includes:
  • WACC
  • Terminal growth rate
  • Guidance: don’t change growth-rate indicators outside the assumptions tab

---

Performance / opportunity framing vs peers (relative value argument)

• The presenter argues Broadcom should get “respect” similar to other AI-related names:
  • Notes AMD and NVIDIA have run
  • AMD described as more expensive
• Claim: analysts’ expectations may be low for AMD, and the presenter questions why Broadcom lacks a similar premium despite comparable growth / FCF potential.
• Time horizon recommendation:
  • Broadcom framed as best risk-reward choice for the next ~3 years

---

CoreWeave / Nebius capacity economics (AI infrastructure forecasting)

The video discusses revenue-per-megawatt forecasts and implied revenue totals for Nebius and CoreWeave.

Key metrics mentioned

Revenue per MW peaks and declines:

• Nebius: peaks at $14.9M per MW in 2029
• CoreWeave: peaks at $14.4M per MW in 2030
• By 2035:
  • CoreWeave: $12.1M per MW
  • Nebius: $11.1M per MW

Capacity (gigawatts) assumptions used:

• Nebius 3.4 GW at the 2029 peak
• Nebius 5 GW by 2035
• CoreWeave 8 GW (stated as an assumed limit in later discussion)

Implied revenue calculations (explicit)

• Nebius peak revenue: $14.9M/MW × 3.4B? (presenter’s math stated as) = $50.66B in 2029
• Nebius 2035 revenue: $11.1M/MW × 5 GW = $55.5B

Interpretation

• Analysts assume only ~10% revenue increase from 2029 to 2035, described as “quite strange.”

Valuation question posed (implied multiple sensitivity)

If Nebius makes ~$50B+ revenue, what multiple applies?

• 5x → ~$250B valuation
• 10x → ~$500B valuation

Caveat

• Assumes revenue is “purely without additional services”; presenter suggests services/software add-ons could justify a premium and higher multiple than a simple P/S framework.

---

Explicit recommendations / cautions

• Implied recommendation: open/expand position in Broadcom
  • Presenter says they invested $15,000 in Broadcom and plan to increase
• Repeated risks:
  • Concentration risk (OpenAI/Anthropic demand)
  • Financing commitment risk if customer growth lags
• Scenario stance:
  • Presenter prefers being conservative to avoid losing money, even if upside may be understated

---

Disclosures / disclaimers

• No explicit “not financial advice” style disclaimer was observed in the provided subtitles.

---

Presenters / sources

• Presenter: single narrator speaker (no name provided in subtitles)
• Tools/resources referenced:
  • fiscal.ai (margin profile link/tool; described with a 15% discount to viewers)
  • Google Drive / Google Sheet for the DCF model (link “below” / “in description” stated)
• Institution mentioned:
  • Bank of America (as one of the banks involved in the described lending/guarantee structure)