Multibagger Selection: Claude Sonnet 5.5's picks
Claude Sonnet 5.5 picked AEHR, LUNR, OUST, EOSE, SERV with a twelve-month thesis, measured against SPY at each checkpoint.
| Experiment | Multibagger Selection |
|---|---|
| Model | Claude Sonnet 5.5 |
| Prompt version | multibagger-selection-v1 |
| Asked | |
| Horizon | 12-month |
| Benchmark | SPY |
| Status | Being measured |
The picks
| Rank | Symbol | Return | Confidence | vs benchmark | Called right |
|---|---|---|---|---|---|
| 1 | AEHR | — | 62% | — | — |
| 2 | LUNR | — | 55% | — | — |
| 3 | OUST | — | 58% | — | — |
| 4 | EOSE | — | 40% | — | — |
| 5 | SERV | — | 30% | — | — |
Checkpoints
| Checkpoint | Portfolio return | Return vs SPY | Best peak multiple | Picks reaching 2x |
|---|---|---|---|---|
| 1 week | -1.22% | +0.72% | 1.18x | 0 |
The model's market view
Small caps had their best first half in decades (Russell 2000 up about 22% by June 30, led by AI-infrastructure names), but the Russell 2000 entered correction territory in mid-October 2026 on rate and inflation worries. Momentum small caps then unwound hard: over roughly 20 trading days, Equibles shows AEHR -42%, EOSE -47%, SERV -40% and OUST -27%. The Fed decides on October 28 and the Russell reconstitution is on December 11. Dispersion is high and equity-financing risk is real, so these picks favour names where a contracted backlog, guided ramp or margin inflection is visible and the price has already reset. All five are very volatile and the evidence quality differs sharply, as the confidence scores show. Several providers disagreed on cash and share data, so conflicts are flagged below.
Why each pick
AEHR
Aehr makes wafer-level and package-level burn-in test systems for AI processors, silicon photonics and power semiconductors. In July 2026 it guided fiscal 2027 revenue to USD 130-150M, up from USD 50.0M in fiscal 2026, with non-GAAP net income of 18-22% of revenue. Effective backlog was about USD 100.6M, covering roughly 67-77% of the guided range. The stock fell about 43% in October from a 52-week high of USD 145.61, which reset the entry valuation while the guidance was left unchanged. I could not verify the fiscal Q1 FY2027 results (reported around early October 2026), so any guidance update since July is data unconfirmed. Reference price is a live intraday quote from Equibles (Oct 27); market cap is the provider's figure at the Oct 26 close.
LUNR
Intuitive Machines has moved from lunar landers toward a broader space-infrastructure prime, adding spacecraft, communications and ground-network businesses through acquisitions (Lanteris, KinetX, Goonhilly, COMSAT). Backlog was a record USD 1.8B at June 30, 2026, versus about USD 300M at end-2025, and bookings were about USD 1.7B year to date. Management reaffirmed 2026 revenue of USD 900M-1B and positive adjusted EBITDA. The stock trades near USD 11 versus a 52-week high of USD 46.75 despite that backlog build. Reference price is an intraday quote at 11:15 AM EDT on Oct 27.
OUST
Ouster sells digital lidar and camera sensors for industrial, robotics, automotive, drone and defence uses. It has beaten or met its revenue guidance in each of the last 8 quarters per Equibles, and gross margin is expanding. It is the highest-quality fundamental profile among the five, with growth above 50%, gross margin above 50% and near-breakeven adjusted EBITDA. The stock fell about 27% in a month and sits far below its 52-week high of about USD 63. The reference price is the latest regular-session close in the provider data (Oct 26); market cap is the provider's figure.
EOSE
Eos makes zinc-based long-duration energy storage systems in the US. Its stock is down about 86% year to date, with market cap near USD 590M against a backlog of about USD 765M and 2026 revenue guidance of USD 300-350M. The thesis is a manufacturing-driven gross-margin inflection from deeply negative toward positive, with the company targeting positive adjusted gross margin by Q4 2026. Financing is the central question: cash is falling, debt is high and short interest is about 38% of float. The reference price is the provider's latest close (Oct 26) and the market cap is the provider's figure at the prior close; the data is volatile.
SERV
Serve Robotics operates sidewalk delivery robots and hospital robots, with more than 2,100 robots deployed. This is the weakest-evidence pick and is kept because it still ranks among the five best candidates I could verify. In August 2026 it cut 2026 revenue guidance to USD 9-10M from USD 26M after Uber Eats volume reversed, and said it does not currently expect to renew the Uber agreement expiring in early 2027. The stock is down about 80% over one year and trades near its liquidity of about USD 240M (as of June 30, 2026), so the bet is asymmetric but depends on revenue diversification and surviving cash burn. It is marginally above the USD 200M market-cap floor and the USD 5M average dollar volume floor (about USD 10M per day). It meets only screen 6.
Other models on the same run
| Model | Picks |
|---|---|
| Gemini 3.6 Flash High | JOBY, RGTI, ACHR, SMR, RDW |
| Muse Spark 1.1 | LEU, LUNR, RDW, EOSE, AMSC |
| Grok 4.7 | CAPR, AGIO, IVA, ACHR, OKLO |
| Perplexity | INOD, LQDA, PGY, MAZE, NVEC |
| GPT-6 | BKSY, AMPX, RCAT, LUNR, SES |
| DeepSeek | INOD, BRUN, ACCV, AIP, LPTH |
ForecastNest provides AI-generated market analysis and quantitative research for informational and educational purposes only. Nothing here constitutes financial, investment, legal, or tax advice.