NVDA$5.3T — LATE
PLTR 3YR+990% — LATE
TSLA301x — LATE
MRVL FY28$18B — NOW
OUST REV+62.8% — NOW
CBRS BACKLOG$25.4B — EARLY
MDB GROWTH+30% — EARLY
NVDA$5.3T — LATE
PLTR 3YR+990% — LATE
TSLA301x — LATE
MRVL FY28$18B — NOW
OUST REV+62.8% — NOW
CBRS BACKLOG$25.4B — EARLY
MDB GROWTH+30% — EARLY
Timing Screen · Nine Names, Three Buckets · 12 September 2026
Early, Right Now, Late
Nine names sorted by where they are in their own cycle — not by whether they are good
9
names · three timing buckets
§ The Screen
3
EARLY — story ahead
of the numbers
of the numbers
3
RIGHT NOW — numbers
arriving, price hasn't
arriving, price hasn't
3
LATE — the move
already happened
already happened
6 of 9
trading below
their 52-week high
their 52-week high
Early
Story is bigger than the revenue. Something real has been built or signed, the money has not landed yet. You are paid to wait — and you carry the risk that waiting turns into never.
Right now
Revenue arrived. Price did not. The rarest of the three, because the proof already exists and you are simply earlier than the crowd.
Late
The easy money is gone. Not bad — harder. A $5 trillion company needs another $5 trillion of buying to double. A $2 billion one needs $2 billion.
Six of the nine are below their 52-week high, several right after record results. When good news gets sold, the bucket matters more than the headline.
§ The Board
All nine, sorted.
Prices as of 12 September 2026.
| Name | Bucket | Size | Growth | From 52w high | What decides it |
|---|---|---|---|---|---|
| ★CBRS | Early | $44B | Rev +103% | −50% | $25.4B backlog converting to revenue |
| MDB | Early | $29B | Rev +30% | — | Growth is re-accelerating, not waiting |
| MP | Early | $9B | Rev +49% | −49% | Robotics demand arriving, not the tech |
| ★MRVL | Right now | $182B | FY28 ~$18B | −13% | Custom silicon doubling in FY28 |
| NOW | Right now | $150B | EPS beat 18% | — | AI agent revenue vs the Iran hit |
| ★OUST | Right now | $2.6B | Rev +62.8% | −45% | Whether the light Q3 guide was a pause or a top |
| NVDA | Late | $5.3T | — | — | Needs another $5T to double |
| PLTR | Late | ~$167 | +990% 3yr | — | Multiple already at the ceiling |
| TSLA | Late | 301x P/E | — | — | 159x even on forward earnings |
§ Early
Early
The story is ahead of the money.
CBRS · Cerebras · $191.84 · half its first-day price
NumbersQ2 2026 revenue +103% year over year. FY26 guidance raised to $880–890M. Backlog of $25.4B, largely tied to OpenAI. Down 17.2% since the last print and roughly half its first-day value.
BullWafer-scale inference silicon with a named anchor customer and a backlog 28 times current annual revenue. Nothing else in the group has that ratio.
BearThat same ratio is the risk. A $25.4B backlog against $885M of revenue is a conversion question, not a revenue figure — and the concentration in one customer means one renegotiation changes the story.
ReadI’d enter with a small position here. The company is trading at roughly 2× the valuation of comparable companies, so I’m not willing to pay full size for it yet. That said, I believe the underlying technology is real.
My view is that we’re simply early. I’d rather wait for the backlog to convert into actual revenue and execution before increasing the position.
My view is that we’re simply early. I’d rather wait for the backlog to convert into actual revenue and execution before increasing the position.
MDB · MongoDB · $373.87 · ~$29B
NumbersQ2 FY27 revenue +30% year over year — the fastest growth in several years, running near $2.8B annualised. The stock fell 13.7% on that print.
BullThe database layer is where AI applications store and retrieve their data, and growth is accelerating, not decaying. Re-acceleration in an established software name is rare and usually gets paid eventually.
BearIt fell on good numbers, which means the multiple is still arguing. Database competition is brutal and the hyperscalers all ship their own.
ReadI’d hold off on jumping in just yet. The revenue is growing and they just turned profitable, so the fundamentals are clearly heading in the right direction. That said, the market is still digesting the valuation.
My view is to stay patient. I’d rather wait to start a position right before it’s ready to take off.
My view is to stay patient. I’d rather wait to start a position right before it’s ready to take off.
MP · MP Materials · $51.36 · 49% off the high
NumbersQ1 2026 revenue +49% to $90.6M, ahead of estimates, with first commercial magnet shipments scaling. A $550M Department of Defense investment and a 10-year price floor of $110/kg on NdPr. Now commissioning heavy rare earth separation — dysprosium and terbium.
BullThose two heavy elements are exactly what keeps a magnet from demagnetising inside a hot robot joint. And the DoD price floor removes the commodity risk that historically made rare earth miners untouchable.
BearHalf off the high for a reason — China retains enormous pricing leverage, and the robotics demand that justifies the premium is still measured in prototypes, not purchase orders.
ReadI’d remain on the sidelines for now. We’re still early on the supply side—capacity is limited, and high-volume demand hasn't quite arrived yet.
My view is that the technology is simply too early in its lifecycle. It’s best to wait for later on when demand catches up before stepping in.
My view is that the technology is simply too early in its lifecycle. It’s best to wait for later on when demand catches up before stepping in.
§ Right Now
Right now
The numbers landed. The price didn't.
MRVL · Marvell · ~$208 · $182B
NumbersRecord Q2 FY27 revenue of $2.74B. Full-year outlooks raised to roughly $12B for FY27 and $18B for FY28. Data-centre growth raised from 50% to about 60%. The stock fell 10% on it.
BullEvery hyperscaler building its own inference chip needs a partner to design it, and Marvell now has a programme with all of them — including a Google agreement worth up to $120B through FY2033. Management said every projection it has made for this category has proven too low.
BearGoogle took warrants for roughly 59M shares as the price of that commitment, and the headline value is milestone-contingent. At a high multiple, the market has stopped paying for contracts and wants delivery.
ReadI’d look to build a position here. The numbers are solid and new orders are actively flying in from across the hyperscaler landscape.
My view is that the window is open right now. It’s time to get in before next year's numbers arrive and get fully priced in.
My view is that the window is open right now. It’s time to get in before next year's numbers arrive and get fully priced in.
NOW · ServiceNow · ~$150B · next print 28 Oct
NumbersQ2 2026 EPS of $0.90 against $0.76 expected — an 18% beat. But Q1 2026 saw the stock sink 14% as subscription revenue took a hit from the Iran war.
BullThe clearest proof in enterprise software that AI is an upsell rather than a threat — agents sold into an existing platform, on top of contracts that already renew.
BearA double from $150B means $300B, which requires the AI agent line to become a genuinely large business, not a feature. And Q1 proved the revenue is not immune to macro shocks.
ReadI’d be a buyer on this setup. The stock just crossed into an uptrend, and demand for AI agents is clearly real considering it grew the most across the entire business.
My view is that the technical momentum and product adoption are aligning right now.
My view is that the technical momentum and product adoption are aligning right now.
OUST · Ouster · $35.33 · $2.55B · 45% off the high
NumbersTTM revenue $204.9M, +62.8%. Q2 revenue of $54.6M beat the $50.9M consensus, up 56%, on 17,000+ sensors shipped. Gross margin 49%. $263M cash, no debt. Analyst target $59.40, roughly 68% above the price.
BullRevenue up 63% while the stock sits 45% below its 52-week high — the cleanest divergence in the group. Design wins keep landing: a I Deere subsidiary, Utah DOT, GeoCue, plus the Stereolabs integration and an NVIDIA relationship.
BearThe crack nobody mentions: Q3 revenue guidance of $54.5–57.5M came in below the $57.48M consensus at the midpoint, and Q2 EPS missed at −$0.27 against −$0.14 expected. Still lossmaking, and the beta is 3.25 — this moves three times the market.
ReadI’d enter a position at these levels. It’s down 45% from all-time highs, and this isn't only a humanoid robotics play—it’s the main publicly listed US lidar pure-play on the market.
My view is that the asymmetric upside is worth taking while the valuation remains discounted.
My view is that the asymmetric upside is worth taking while the valuation remains discounted.
Nefarious Trading
Equity research and trading commentary — AI infrastructure, semiconductors, aerospace, energy, commodities.
Sources
Prices, market caps and 52-week ranges as of 11–12 September 2026. Company figures from the most recent reported quarters: Cerebras Q2 2026, MongoDB Q2 FY27 (quarter ended 31 July 2026), MP Materials Q1 2026, Marvell Q2 FY27 (27 August 2026), ServiceNow Q2 2026 (22 July 2026) and Q1 2026, Ouster Q2 2026 (6 August 2026). The MP Materials Department of Defense partnership terms — $550M and a 10-year $110/kg NdPr price floor — are from the company's own announcement. The Marvell–Google agreement value of up to $120B through FY2033 is milestone-contingent per Marvell's disclosure. Palantir's three-year cumulative figure is derived from a reported 122% three-year average annual return and is marked as calculated, not reported. ServiceNow per-share quotes conflicted across sources on 12 September, so only market cap is used.
Whose opinion is whose: the EARLY / RIGHT NOW / LATE assignments are my. The supporting figures are public company disclosures, date-stamped above; the starred rows and the bucket commentary are my reading of that data.
Prices, market caps and 52-week ranges as of 11–12 September 2026. Company figures from the most recent reported quarters: Cerebras Q2 2026, MongoDB Q2 FY27 (quarter ended 31 July 2026), MP Materials Q1 2026, Marvell Q2 FY27 (27 August 2026), ServiceNow Q2 2026 (22 July 2026) and Q1 2026, Ouster Q2 2026 (6 August 2026). The MP Materials Department of Defense partnership terms — $550M and a 10-year $110/kg NdPr price floor — are from the company's own announcement. The Marvell–Google agreement value of up to $120B through FY2033 is milestone-contingent per Marvell's disclosure. Palantir's three-year cumulative figure is derived from a reported 122% three-year average annual return and is marked as calculated, not reported. ServiceNow per-share quotes conflicted across sources on 12 September, so only market cap is used.
Whose opinion is whose: the EARLY / RIGHT NOW / LATE assignments are my. The supporting figures are public company disclosures, date-stamped above; the starred rows and the bucket commentary are my reading of that data.
One trader's view — not investment advice. Do your own research. Bucket calls are opinion; every figure is date-stamped above.