Built by Claude FableInvestor Sentiment (Inverted) (7/10)
Bottom line: This is a real management-versus-street gap, and it sits in AI. Management is loudly and specifically bullish on data center AI inference ($5B FY27, $15B FY29) and on a $40B FY29 non-handset business. The sell side is mostly unconvinced: Buy ratings have fallen from about 55% to about 30% of coverage, and price targets were cut after FQ3'26. Most of the doubt is aimed at data center, which today brings in less than $100M a quarter. What keeps this short of an 8 to 9: no insider has bought stock in the open market since 2019, the CEO and CFO sold steadily all through 2026, and Qualcomm has tried data center twice before without success (Centriq server CPU, Cloud AI 100).
What management is saying (FQ3'26 call, 2026-07-29, and Investor Day, June 2026)
- Data center revenue target of $5B in FY27 and $15B in FY29. Total non-handset target raised to $40B by FY29. The auto plus IoT part went from $22B to $24B, with auto alone now $10B.
- "We have POs from these engagements, and so we've already started wafers" (Akash Palkhiwala). The 2 hyperscaler custom silicon programs start producing revenue in the December quarter.
- FY27 non-handset growth is expected to accelerate from 24% to more than 60% y/y, which would replace all of the Apple product revenue that drops out.
- HBC Gen 1 accelerator has taped out. Silicon demos are coming in the next few quarters and the product launches mid-2027. Amon: "in active conversations with nearly every leading data center player."
- Amon also said this directly: "we recognize that investors want to see more proof points that we can successfully execute on our plans as a new entrant." Management is openly saying the street does not believe it yet.
The gap: management's target versus reported numbers
| Metric | FQ3'25 | FQ2'26 | FQ3'26 | y/y | Mgmt forward claim |
|---|---|---|---|---|---|
| Data center revenue | n/a | $97M | $88M | n/a | $5B FY27 (about 14x the current run-rate) |
| QCT Automotive | $984M | $1,326M | $1,588M | +61% | about $7B annualized exiting FY26 (raised from $6B) |
| QCT IoT | $1,681M | $1,726M | $1,830M | +9% | Industrial: $8B FY29, more than $7B pipeline |
| QCT Handsets | $6,328M | $6,024M | $5,086M | -20% | China bottomed in FQ3. Android to grow 5% CAGR once memory normalizes |
| Total revenue | $10,365M | $10,599M | $9,947M | -4% | Top-line growth in FY27 |
Street estimates versus management's framing (FMP consensus averages)
| FY26E | FY27E | FY28E | FY29E | |
|---|---|---|---|---|
| Street revenue | $43.00B | $45.04B (+4.7%) | $52.37B (+16%) | $62.84B (9 analysts) |
| Street EPS | $10.54 | $10.13 (-4%) | $13.07 | $17.65 (2 analysts) |
| Mgmt framing | Non-handset +24% | Non-handset more than 60% (about $20B+), DC $5B | n/a | Non-handset $40B, DC $15B |
What the table shows:
- The street has taken the FY27 revenue direction but not the earnings payoff. Consensus FY27 EPS is still below FY26. Analysts are modeling the Apple loss, about 1.5 to 2 points of gross-margin dilution from custom ASICs, and OpEx growth from Alphawave and Modular, but are giving little credit for data center operating leverage.
- The FY28 and FY29 consensus is so thin (9 revenue estimates and 2 EPS estimates for FY29) that most of the street has effectively not modeled the $15B data center target. That is the clearest sign the AI case is not being underwritten.
What analysts asked about on the FQ3'26 call (8 questioners). The questions were defensive and asked for proof, not upside:
- Margins and pricing (TD Cowen, JPMorgan, Bernstein): when the input-cost pass-through shows up, and whether data center dilutes gross margin.
- Apple (Bernstein's Rasgon): "are you basically starving Apple?"
- Data center (TD Cowen, Melius): what the ramp looks like and whether there are customers "beyond the 2." Amon's reply: "everybody wants instant gratification... customers want to see silicon."
- China and restrictions (BofA, Melius): whether the China bottom will hold, and US-China risk to data center.
- Nobody asked about AI PC, and only management raised the agentic-handset cycle. The street is not engaging with the edge-AI side of the story at all.
How management's message has built over time (counts of "data center" in each transcript): 0 (FQ4'24), 1 (FQ1'25), 0 (FQ2'25), 11 (FQ3'25, Alphawave and first hyperscaler talks), 17 (FQ4'25, AI200/HUMAIN), 17 (FQ1'26), 19 (FQ2'26, custom silicon win), 20 (FQ3'26, $5B/$15B targets). The message has been consistent and has grown more specific each quarter. It has moved from "advanced discussions" to a signed hyperscaler, then POs, then wafers started. That is the pattern of a management team building a case before the street believes it.
Spending and hiring backing the claim
- R&D is growing in the mid-teens while revenue is falling 4%, capex is up 69% y/y (FY26 YTD $1,578M, about 2x the FY25 first-three-quarters pace), and buybacks were halved in FQ3 while the company paid for Alphawave, Ventana and Modular. Management is putting money behind the data center plan. That makes the stated conviction more credible.
Management track record on promises like this
- Delivered: auto. FQ3'26 auto revenue grew 61%, the exit run-rate target went from $6B to $7B, and the FY29 auto target rose to $10B. Auto guidance has been met or beaten (FQ3'26 guided +60% low end). FQ3'26 revenue came in at the high end of guidance. Management also flagged the Apple step-down years in advance.
- Weak spot: data center. Centriq (2017) was abandoned and Cloud AI 100 never reached scale. The street's skepticism here has history behind it, not just bias. The FY27 custom-ASIC revenue is backed by POs, so that piece is about execution. The FY29 merchant accelerator and CPU revenue is still unproven.
| Measure | Nov 2024 | Jan 2026 | Oct 2026 |
|---|---|---|---|
| Strong Buy + Buy | 21 | 17 | 11 |
| Hold | 16 | 20 | 23 |
| Sell + Strong Sell | 1 | 1 | 3 |
| Buy share | 55% | 45% | 30% |
Source: FMP grades-historical. A broader FMP snapshot shows 1 Strong Buy, 29 Buy, 34 Hold and 5 Sell, with consensus "Hold".
- Recent actions were mostly negative: downgrades from Mizuho (Jan), Seaport to Sell (Mar), Bernstein, JPMorgan and BNP (Mar/Apr), and Piper initiating at Neutral (Sep). Barclays and BofA are Underweight/Underperform. The only upgrades were Morgan Stanley (Underweight to Equal Weight), Daiwa and Loop.
- After the FQ3 print (2026-07-30), 8 of 9 target changes were cuts. Examples: RBC $250 to $160, Wells $265 to $170, Bernstein $235 to $165, Barclays $245 to $180. The only exception was Baird, which raised to $400. That is a single believer against a skeptical pack.
- The average target is $204.48 (median $190, range $120 to $400) against the FMP quote of $174.71 (2026-10-08), about 17% above the current price (median about 9%). The spread from $120 to $400 is unusually wide for a mega-cap and confirms the street has no shared view on the AI story.
| Period | Open-market buys | Notable sales |
|---|---|---|
| FY2026 YTD (Jan to Oct) | 0 | CEO Amon sold 40,000 sh in 10,000-share blocks (May 4/5, Sep 21/25), apparently at preset prices of $180 to $200. CFO Palkhiwala sold every month (Apr to Sep). Chief HR Officer sold 6,400 sh |
| Last open-market purchase by any insider | 2019 | n/a |
Source: FMP insider-trading (Form 4). The sales look like scheduled, routine selling: fixed block sizes at round-number limit prices. They are not a red flag on their own. However, there has been no open-market buying in 7 years, including at the 52-week low near $122. That gives zero insider confirmation of the data center thesis. On this rubric factor, insiders count against the score.
- Stocktwits followers are about 78k, up 15% over the 3 months to mid-2026. Retail mood on Stocktwits and Reddit swung from about 85/100 around the June Investor Day to "fear" (about 34/100) in July after the print. Attention is moderate and reactive, not crowded.
- Short interest was 4.3% to 5.4% of float (May to July 2026) and the 13F put/call ratio was 1.14 (June 2026). Both are bearish positioning for a ~$183B company.
- 13F holders rose by 324 to 3,185 in the June quarter, with 467 new positions. Institutions are building positions around the Investor Day, but the ratings show they are not convinced. QCOM is also held in 5 Autopilot leaderboard portfolios (avg weight 0.151), which is small.
| Factor | Weight | Read | Sub-score |
|---|---|---|---|
| Mgmt-Street divergence | 50% | Clear and specific on data center AI. Mgmt says it outright. Street FY27 EPS below FY26 and FY29 barely modeled. Spending backs it up | 8.5 |
| Analyst skepticism | 20% | Buy share fell from 55% to 30%, 8 of 9 post-print targets cut, consensus Hold | 8 |
| Track record (applied to divergence) | inside divergence | Strong in auto and on guidance accuracy. Weak history in data center | lowers divergence from 9 to 8.5 |
| Insider activity | 15% | No buys in 7 years. CEO and CFO selling on a schedule | 3 |
| Retail attention | 15% | Moderate, reactive, short interest elevated | 6.5 |
| Weighted | 7.2, rounded to 7 |
Assessment. On the thesis question of whether anything changed in FQ3'26: yes, and the change widened the gap between management and the street rather than closing it. The same quarter that brought harder AI proof points (custom-silicon POs and wafer starts for 2 hyperscalers, HBC tape-out, the Modular deal closed, auto exit run-rate raised to $7B, a $5B FY27 data center target) also brought 8 of 9 target cuts, a shift toward Hold ratings, and a consensus that still has FY27 EPS below FY26. The street is focused on the visible negatives: the Apple exit, the memory-driven handset decline (-20% y/y), gross-margin dilution and higher OpEx. It is giving almost no value to a business management says goes from $88M a quarter to $5B a year within 12 months. That is the "management is the contrarian" setup the framework looks for, and it is not just a cheap, disliked stock: management keeps naming specific, checkable milestones, starting with December-quarter data center revenue. The setup falls short of a 9 or 10 for two reasons. Insiders are not buying (they are selling steadily), and data center is the one area where Qualcomm's own history supports the skeptics. The FQ4'26 print (early November) and the FQ1'27 data center revenue line are the tests. If custom-silicon revenue shows up on schedule, the street will have to start believing.
Data sourced from Daloopa. Consensus, targets, ratings, insider and 13F data from FMP (/stable). Retail and short-interest data from web search: Benzinga short interest, StreetBeat, AltIndex Stocktwits, AltIndex sentiment, CFGI monthly.