The raw quarterly tape is the most informative thing in this file, and it's louder than any model output: revenue went $14.7M → $60.7M → $63.9M → $52.9M → $38.0M → $64.2M → $126.6M → $126.6M. The sequential doubling from Aug to Nov 2025 followed by a flat Feb 2026 print is the entire story — the Ellendale HPC contract appears to be ramping, but the flat sequential is the first yellow flag that the ramp may be lumpier than bulls think. Meanwhile the Feb 2026 net loss exploded to -$99.3M on the same revenue base as Nov's -$17.5M, which means operating leverage went the wrong way quarter-on-quarter — likely depreciation/interest from the GPU buildout hitting the P&L, but worth scrutinizing before assuming the margin trajectory bends positive. Annual FCF of -$797M against $113.9M cash means this company has already raised, and will continue to raise, very large amounts of capital. The balance sheet tile is missing total debt — a critical omission given a $681.6M annual capex run rate funded externally.

On the prior models: the synthesis verdict of "High Conviction Required" is directionally correct but soft. At $12.7B market cap on $215M trailing revenue (59x P/S) and roughly $500M annualized run-rate from the last two quarters (~25x forward P/S), the stock is priced for CoreWeave-like execution without CoreWeave's $20B+ contracted backlog visibility in this file. Market Forces' "avoid" is harsher and, I think, closer to right on risk-adjusted basis — the model correctly flags that GPU-as-a-Service is shifting from scarcity to potential oversupply as Oracle, Microsoft, AWS, and CoreWeave all add capacity through 2026. The Pre-Flight thesis implicitly requires 2-4% AI training share by 2028 — that is an aggressive ask for a sub-scale operator competing against hyperscalers with cost-of-capital advantages of 300-500bps. The "Adequate Revenue Confidence" tag is too generous given one flat sequential quarter; "accelerating" is a backward-looking artifact.

The contrarian bull case I'd steelman: Ellendale is a single-tenant hyperscaler contract (reportedly CoreWeave/Microsoft adjacent) with 10-15 year duration, and once steady-state, this becomes an infrastructure annuity at 60-70% gross margins, not a 10.5% gross margin business. If you annualize the $126.6M quarterly run-rate at maturity with proper scaling, $1.5-2B revenue with 25-30% EBITDA margins gets you $400-600M EBITDA, and at 20x that's $8-12B EV — i.e., the stock is roughly fair to slightly rich today, not 3x overvalued. But this requires (a) the contract converting fully without renegotiation, (b) no further dilution beyond what's modeled, and (c) the gross margin inflection actually showing up in the next 2-3 prints. The insider activity — multiple M-Exempt + F-InKind + open-market sales in April 2026 — is exactly the option-exercise-and-flip pattern you'd expect at the top of a narrative move, not insiders accumulating. That's a real tell.

I dissent partially from the synthesis. "High Conviction Required" implies the bull case is live with proper risk management; I'd argue the asymmetry is worse than that framing suggests because the dilution risk is unquantified in this file (no debt figure, $797M FCF deficit, 0.77 current ratio). At $44.45, you're paying for flawless execution in a market where the customer concentration risk is enormous and the moat is essentially "we built data centers near cheap power" — replicable. Fair value under base-case execution is closer to $25-30 (15-18x mature EBITDA, discounted for dilution and 3-4 year wait); the bull case to $60-70 requires multi-tenant expansion beyond the lead contract. I would not initiate here; I'd want either a pullback to the $28-32 range or a clean Q4 print showing gross margin moving from 10% toward 25%+ with no surprise raise.