Six months ago the market left Qualcomm for dead — a phone-chip company you trade around the handset cycle and forget. Then it doubled the FY29 target, named Meta as a CPU customer, and built a data center out of a slide. The story is now real. The one number that decides the re-rate is the one they wouldn't show.
First gate: where's the trend? Honestly — mixed. A big run, a sharp pullback, and a catalyst that faded.
I have to be straight here, because this one doesn't pass the gate as cleanly as the bull tweets suggest. Yes, the multi-month uptrend off the March lows is intact — QCOM ran from ~$122 to ~$260, a near-double on data-center anticipation. But it's since pulled back ~21% from that high, and the Investor Day itself was a "sell the news" tape: the stock fell ~4% during the session on June 24, popped ~10% after-hours on the announcements, gapped open near $219 the next morning — and then faded back to ~$205. That's not the clean, order-backed grind I want to see; it's a high-volatility consolidation where the marquee catalyst couldn't hold a bid. The trend gate reads amber, not green: the long-term direction is up, but the immediate momentum rolled over right when the story got its loudest. This is a "prove it" zone, not a "chase it" one.
The stock nearly doubled this year, then dropped ~21% from its peak. On the big Investor Day, it actually fell during the day, jumped after hours, then gave most of that jump back. So the long-term trend is up, but right now it's choppy and the exciting news didn't make it stick. That's a yellow light, not a green one.
A toll booth wearing a phone-chip costume. Two great businesses, one fading one, and a brand-new bet.
Qualcomm is really three things: a patent business that prints money (a "toll booth"), a chip business where phones are shrinking but cars and gadgets are growing, and a brand-new data-center business that just went from a dream to a real plan with real customers. The market mostly prices the phones and ignores the rest.
They doubled the target. $40B non-handset, $15B data center, $18+ EPS by FY29.
| FY2029 target | Prior | New | Read |
|---|---|---|---|
| Non-handset (QCT) revenue | ~$22B | ~$40B | roughly doubled |
| Data center revenue | ~$0 | $5B FY27 → >$15B FY29 | the number that moved it |
| Automotive revenue | — | $10B | $65B design-win pipeline (was $45B) |
| IoT revenue | — | >$14B | edge / PC / industrial / robotics |
| Handset mix (FY29) | majority | ~1/3 of revenue | truly diversified |
| Non-GAAP EPS | ~$15.50–16 (bull model) | >$18 | above what bulls penciled |
| Data-center gross margin | — | not disclosed | the missing number |
For context on why >$15B mattered: Bank of America went into the day arguing even $10B of data-center revenue by 2028 was already priced in. Management guided well above that bar — with a nearer-term $5B data-center waypoint in FY2027 ramping to >$15B by FY2029 — and stapled real product and a named hyperscaler to it. The same day, Qualcomm announced it's acquiring AI-software startup Modular for ~$3.92B (founder Chris Lattner; the Mojo/MAX stack), the move that attacks NVIDIA's real moat — software — not just the silicon. CEO Cristiano Amon's headline framing: by FY2029, handsets shrink to just one-third of revenue. On paper, this is the cleanest "we're not just a phone company" statement Qualcomm has ever made — the market's muted, fade-y reaction tells you the skepticism is about quality and timing, not the headline numbers.
Qualcomm doubled its five-year sales goal and said it'll earn more than $18/share by 2029, driven by a new $15B data-center business that didn't exist before. It even bought a software company to take on NVIDIA where NVIDIA is strongest. Big, credible goals — but the market wants proof, not promises.
Does the royalty survive Apple? The chip sale dies. The patent license is a separate — and expiring — question.
| The toll booth | The detail |
|---|---|
| QTL economics | ~$1.4B/quarter at ~72% EBT margin — software-like profitability on 5G royalties |
| Apple's weight | Apple is roughly ~50% of QTL — the single biggest concentration risk in the whole company |
| The chip vs the license | Apple's in-house modem (C1/C2) takes Qualcomm's chip-supply share toward zero — but the patent royalty is a separate license on 5G IP Apple still uses |
| The real cliff | The Apple license has been extended to March 2027. The threat isn't the modem — it's the renewal terms after 2027 |
This is the heart of the bull-bear fight, so let's be precise. The bear shorthand — "Qualcomm dies when Apple builds its own modem" — conflates two different things. The chip sale to Apple is indeed going away. But the patent royalty is a separate toll: even with an Apple-made modem, Apple's phones use 5G standards that read on Qualcomm's patents, so the royalty has historically survived the chip loss. That's the "toll booth survives Apple" core. The honest caveat: the license runs only to March 2027, and Apple — now with its own silicon and less leverage to lose — will negotiate hard. The toll booth probably survives; the rate may not survive unscathed. Watch the renewal, not the teardown.
People think Qualcomm collapses when Apple makes its own modem chip. But Qualcomm gets paid two ways from Apple: selling the chip (going away) and charging a patent fee on 5G (separate). The patent fee usually survives even without the chip — but the current deal expires in March 2027, and Apple will push for a lower rate. So the toll booth likely lives; the size of the toll is the question.
The socket-and-hold playbook, aimed at AI. Real product, real customers, real software — and the LPDDR angle.
The strategic reframe is the real story: Qualcomm wants the industry to stop measuring raw FLOPS and start measuring tokens-per-watt. The bet is that as agentic AI and "query swarms" explode an estimated 40x between 2026 and 2030, the binding constraint becomes power and compute economics, not peak performance — exactly where 20 years of mobile low-power design is supposed to pay off. It's a genuinely differentiated pitch. It's also, for now, a claim: 8x tokens-per-watt is a slide until independent silicon proves it in production.
Qualcomm's data-center chip skips the pricey "HBM" memory rivals depend on, which could make it cheaper and more power-efficient — its whole pitch, measured in "tokens per watt" rather than raw speed. It landed Meta as a multi-year customer (plus Microsoft, ByteDance, a Saudi project) and bought software to challenge NVIDIA's biggest advantage. The catch: the flagship CPU doesn't ship until ~2028 (though some revenue starts in 2027).
The one number they wouldn't show. Revenue targets are concrete. Margin quality is a black box.
Management gave revenue targets and an EPS target — but refused to put a clean gross-margin number on the data-center business. That's not a footnote; it's the whole investment case. Qualcomm today earns a premium multiple because its mix is high-margin (72% licensing, fat handset/auto chip margins). If $15B of data-center revenue arrives at commodity hardware margins — competing on price against NVIDIA, AMD, and the ASIC crowd — then the revenue grows but the blended margin falls and the premium multiple compresses. Growth without margin is a re-rating trap. If, instead, the LPDDR/power-efficiency edge makes Dragonfly genuinely differentiated, the dollars are high-quality and the stock re-rates. Same $15B; opposite outcomes for the stock. Until they disclose it, the bull case on revenue is strong and the only question that matters is unanswered.
Qualcomm told us how much data-center money it expects, but not how profitable it'll be. That matters more than the revenue: cheap, low-profit sales would actually drag the company's overall quality down and could hurt the stock even as sales grow. High-profit sales would lift it. They didn't say which — so the most important question is still open.
The platform is wider than servers. A $65B auto pipeline, physical-AI robotics, and 6G on deck.
Qualcomm isn't only chasing data centers. Its car-chip order book grew to $65B (Stellantis and others), it's moving the same tech into robots ("physical AI"), and it's already designing 6G — which keeps the patent toll booth alive into the next decade. The "edge everywhere" story is broader than the headlines.
Walking into NVIDIA's house. Qualcomm's angle is inference, power, and price — not training.
| Player | Stronghold | vs QCOM |
|---|---|---|
| QCOM · Qualcomm | Power-efficient inference CPU + LPDDR, edge AI | Subject — late entry, differentiated angle, unproven at scale |
| NVIDIA | AI training + the CUDA software moat | The 800-lb gorilla; QCOM avoids training, attacks inference/TCO |
| AMD / Intel | x86 server CPUs, AMD MI accelerators | Incumbents in the CPU socket QCOM wants to take |
| Broadcom / Marvell | Custom hyperscaler ASICs | Direct rivals for the "custom silicon for a hyperscaler" deals |
| Ampere / AWS Graviton | ARM server CPUs already deployed | Prove ARM-in-datacenter works — both tailwind and competition |
Two things can be true. The power-efficiency thesis is real: inference and always-on AI care about watts-per-token and total cost of ownership, and Qualcomm's 20 years of mobile low-power design plus a no-HBM architecture is a genuine angle as the industry hits the "power wall." But the execution bar is brutal: NVIDIA's moat is software (CUDA), not just chips; ARM-in-datacenter is still proving itself; the Dragonfly flagship doesn't ship until ~2028; and Qualcomm has tried the data center before — its Centriq server effort died in 2017. The Modular acquisition shows they learned the lesson (software first), but "credible angle" and "winning share from NVIDIA" are very different sentences.
Qualcomm isn't trying to beat NVIDIA at training AI — it's targeting the cheaper, lower-power "inference" side where its mobile expertise helps. That's a smart, real angle. But NVIDIA's true lock-in is its software, the flagship chip is years away, and Qualcomm already failed at servers once (2017). Good plan; very hard game.
The price ran ahead of the analysts. Hold consensus, ~$178 target — below the stock.
| Gauge | Reading | Read |
|---|---|---|
| Consensus rating | Hold (~30–35 analysts) | skeptical, not euphoric |
| Average price target | ~$178 | below the ~$205 price |
| Target range | ~$100 → ~$300 | huge dispersion = low conviction |
| Bull anchor | JPMorgan $265 (Positive Catalyst Watch) | the re-rate case |
| Cautious anchor | RBC $175 ("awaiting DC progress") | wants proof first |
| Forward P/E | ~17–18x NTM EPS (~$11–12) | cheap for "AI," normal for QCOM |
| Dividend | ~1.8% yield + buyback | you get paid to wait |
Here's the paradox that makes QCOM interesting: the sell-side average target (~$178) sits below the current price, and the consensus rating is a Hold. On one reading, that's the bull's dream — Wall Street hasn't repriced the data-center story, so there's room for upgrades as it de-risks. On another, it's a warning — the price has already run past where the analysts think it belongs, which is exactly why a validated Investor Day still produced a fade. Both are true. At ~17–18x forward earnings with a real dividend and buyback, you're not overpaying for the legacy business; you're paying a modest option premium for a data-center call that the Street, by its own targets, doesn't yet believe.
Most analysts rate it a "Hold" and their average price target (~$178) is actually below today's price (~$205) — meaning the stock has run ahead of what the pros think it's worth. That can mean either "upgrades are coming once it proves itself" or "it's gotten ahead of itself." It's cheap on normal earnings and pays a dividend, so you're paid to wait for the answer.
A genuinely two-sided setup. Cheap optionality vs an unproven, undisclosed-margin pivot.
▲ The Bull Case
- 72%-margin toll booth the market still underrates
- $40B FY29; DC $5B FY27 → $15B FY29; $18+ EPS
- Meta multi-gen CPU + Microsoft, ByteDance, Humain
- 8x tokens/watt (claimed) — LPDDR/no-HBM inference edge
- Street at Hold / $178 — room to upgrade as it de-risks
- Record auto ($6B+ run-rate) + buyback + 1.8% yield
▼ The Bear Case
- DC gross margin undisclosed — the core question unanswered
- Apple ~50% of QTL, license expires March 2027
- Already ran ~2x; the pop faded — easy money's gone
- CPU flagship is 2028; FY27 rev leans on accelerators
- CUDA moat; Qualcomm's 2017 Centriq server flop
- Handset secular decline + Android memory-price pressure
Bull: a money-printing patent business plus a credible, customer-backed new growth engine that's cheap and under-followed. Bear: the new business's profitability is a mystery, the Apple deal is up for renewal in 2027, the stock already doubled, and beating NVIDIA is brutally hard. Both arguments are strong — which is why this is a debate, not a slam dunk.
My 12-month scenarios. Re-rate, grind, or de-rate — and the margin number is the fork.
The fork is unusually clean. If Qualcomm discloses (or proves) healthy data-center margins and adds another U.S. hyperscaler while renewing Apple on tolerable terms, the Hold-rating crowd capitulates and the stock re-rates toward the bull anchor — that's the $280–320 path. If the data-center dollars look like commodity hardware, or the Apple 2027 renewal cuts the toll, the premium multiple compresses back toward the ~$178 consensus. The base case — and where I'd put the most weight — is a "prove it" grind: the targets are real enough to keep a bid, but without the margin number the market makes Qualcomm earn the re-rate one quarter at a time. You're not buying a breakout here; you're buying a multi-quarter show-me story with a dividend while you wait.
Three paths over the next year. If Qualcomm shows the new business is profitable and renews Apple well, the stock could run to ~$280–320. If the new business looks cheap-and-cutthroat or Apple cuts the fee, it could fall to ~$150–175. Most likely it grinds in between (~$210–250) as it proves itself bit by bit. The profit-margin reveal is the deciding factor.
Two scores. The business: 7.5/10 · the entry at ~$205: 6/10.
| Dimension | Grade | Why |
|---|---|---|
| Toll booth (QTL) | A− | 72% margins, global 5G royalty — minus the Apple/2027 overhang |
| Auto / IoT | A− | Record auto, $6B+ run-rate, proven socket-and-hold |
| Data-center optionality | B | Real product + Meta, but undisclosed margin & 2028 ship |
| Valuation | B+ | ~17–18x with dividend + buyback; cheap for the optionality |
| Handset core | C | Secular decline + Apple chip loss + Android pressure |
| Entry timing | C+ | Ran ~2x, pop faded, ~$178 consensus below price |
The gap between 7.5 and 6 captures it: a genuinely good, genuinely mispriced business whose cleanest mispricing already got partially corrected by a near-double off the lows. I agree with the core thesis — the toll booth survives Apple, the data-center pivot is real, and the Street is behind it. I just won't pretend the layup is still sitting at $122. At $205, into a faded catalyst with the one decisive number (DC margin) still hidden, this is a build-it-on-weakness, prove-it-quarterly name, not a chase. The dividend and buyback mean you're paid to be patient — so be patient.
The company is a 7.5/10; buying it right now at ~$205 is a 6/10. It's a good, under-appreciated business — but the obvious bargain was at $122, and the recent excitement already faded. Better to add on dips and let it prove the profit story, collecting the dividend while you wait, than to chase it here.
John's read. The thesis is right. The discipline is to wait for the one number.
- The core thesis holds up, and Investor Day validated it. The market does price QCOM like a dying phone-chip company, and that's wrong — there's a 72%-margin toll booth, a record auto business, and now a customer-backed data-center plan underneath the label. The "mispriced large cap" framing is fair.
- But the easy money was at $122, not $205. The stock already nearly doubled on this exact anticipation, and the Investor Day pop faded the same day. Repricing has started — the layup is gone. What's left is a show-me story, and I'm going to price my entry accordingly.
- The toll booth survives Apple — the rate is the risk. The chip sale to Apple is dead; the patent royalty is separate and historically survives. The thing I'm actually watching is the March 2027 license renewal on ~50% of QTL, not the modem teardown headlines.
- The whole re-rate hinges on one number they wouldn't show. Data-center gross margin. If those $15B of dollars are high-quality, this re-rates toward $280+. If they're commodity, revenue grows and the multiple compresses anyway. Until management discloses it, I treat the bull case as unconfirmed, however good the revenue targets look.
- How I'd play it: this is a build-on-weakness, get-paid-to-wait name — not a breakout chase. I'd accumulate on pullbacks toward the low-$180s/$190s (where the Street's target lives), keep it sized as a thesis position not a trade, and key adds to hard proof points: the DC margin disclosure, the $5B FY27 data-center ramp actually showing up, a second U.S. hyperscaler, the Apple 2027 renewal, and the late-July earnings print. Right thesis, wrong price to chase — let it come to you.
Want the proof points that turn the QCOM grind into a re-rate?
Join the Discord to find out! →Investor Day (June 24, 2026, New York — livestream) — FY29 targets (non-handset ~$22B→~$40B; data center $5B FY27→>$15B FY29; auto $10B with design-win pipeline $45B→$65B; IoT >$14B; non-GAAP EPS >$18; ~$1.7T TAM by 2030; handsets fall to ~1/3 of revenue by FY29), Dragonfly platform / C1000 CPU (250+ core, >5GHz, PCIe Gen7, CXL, LPDDR/no-HBM, avail. ~2028) + AI200/AI250 accelerators, the "tokens-per-watt" reframe (claimed up to 8x vs GPU, 200x+ memory vs SRAM; query-swarm growth ~40x 2026→2030), meaningful DC revenue from fiscal Q1 2027, Meta multi-gen CPU (deploy ~H2 2028), Microsoft Azure (HBC), ByteDance, Saudi Humain (~1.9GW), Stellantis auto, physical-AI / robotics (System 0/1/2), 6G roadmap (CEO Cristiano Amon), Modular acquisition (~$3.92B): Qualcomm Investor Day 2026 livestream & IR, CNBC, Yahoo Finance, Seeking Alpha, ServeTheHome, TipRanks. Financials — Q2 FY2026 (ended Mar 29, 2026): revenue $10.6B, non-GAAP EPS $2.65; QCT $9.1B (handset $6B, IoT $1.7B +9%, record auto $1.3B +38%, >$5B annualized → >$6B exit run-rate); QTL $1.4B at 72% EBT; $3.7B returned ($2.8B buybacks + $945M dividends); Q3 guide $9.2–10B rev / $2.10–2.30 EPS: Qualcomm Q2 FY26 earnings release & 8-K, Motley Fool transcript, StockTitan. Licensing/Apple — Apple ~50% of QTL; 5G patent license extended to March 2027: Tom's Hardware, Futurum, exploresemis. Valuation/Street — consensus Hold, avg target ~$178 (range ~$100–$300), JPMorgan $265 (Positive Catalyst Watch), RBC $175; QCOM −4% on Investor Day (regular session): MarketBeat, Public.com, Barchart, TradingKey. Market data — IBKR (last ~$204.71, +3.7%; prior close $197.41; intraday high $219.43; 52W $121.99–$259.92; YTD +20.5%; IV ~71%, 89th pctile; div yield ~1.82%).