Corvex arrives at its September-quarter report with a stock that has been cut in half since August and a business story that, on paper, has only gotten better. That gap between operating momentum and market verdict is the whole setup. The former Movano Health, now a GPU cloud provider after its March reverse merger, reports before the open on November 12. This report is less about a headline beat than about proving the company can turn contracted compute into revenue fast enough to stay ahead of its funding needs.
Conventional earnings yardsticks offer little help here. The year-ago quarter reflected a smart-ring company generating roughly $0.1 million in revenue and losing $4.73 per share, so any year-over-year comparison is essentially comparing two different companies. The more useful benchmark is internal. Last quarter Corvex booked $3.8 million in AI platform revenue and introduced a new metric, roughly $22 million of contracted annualized recurring revenue on live compute, which included capacity delivered on August 4. Divide that figure by four and you get a quarterly pace of about $5.5 million. With no formal guidance from management, revenue landing meaningfully above the prior $3.8 million and moving toward that implied run rate would be the cleanest validation that the Blackwell expansion is live and billing.
The quality of that revenue matters as much as the amount. Management stressed that every contract is take-or-pay, fixed-term, and underwritten to a project-level return hurdle, and that demand exceeds supply, with power as the binding constraint. If that framing holds, the ARR figure should rise as new capacity comes online rather than drift lower. Management did leave itself an escape hatch, though, warning that ARR is a point-in-time number that could dip temporarily as capacity shifts toward the Token Factory inference product, now in closed alpha, before that product's economics match bare-metal rentals. A small, clearly explained dip tied to Token Factory progress would be tolerable. An unexplained decline, or more reliance on a handful of already concentrated customers, would undercut the demand narrative.
Profitability and cash are where the bear case lives. AI platform adjusted EBITDA worsened to a $2.3 million loss last quarter from a pro forma loss of about $0.9 million, and the company posted a $12.8 million net loss. Cash slid to $21.7 million from more than $29 million a quarter earlier, and management flagged significant additional capital needs. The Blackwell buildout was financed with a mix of debt, customer prepayment and cash, which means the balance sheet that showed no funded debt after the bridge loan was retired is likely changing. Investors should look for EBITDA losses narrowing as new GPUs absorb fixed costs, and for a concrete financing plan from new CFO Chance Moreland, the former Morgan Stanley banker brought in to build the capital formation engine. Progress on unremediated internal control weaknesses would also help credibility.
The market has clearly chosen to focus on those risks. Shares sit at $5.26, a hair above the post-report low of $5.12 and far beneath the 200-day moving average near $13.91, after trading as high as $26.72 in the prior quarter. A resale registration covering up to 53.4 million shares and roughly $143.6 million of stock compensation running through 2030 add a persistent dilution overhang, and sentiment reads roughly 73% bearish. That pessimism lowers the bar for a relief move, but it also signals that investors expect another equity raise.
Ultimately, the report hinges on unit economics against the clock. If revenue tracks toward the contracted run rate, losses start to narrow, and new capital arrives on terms that do not swamp existing holders, the neocloud pivot looks real and the selloff looks overdone. If growth stalls while cash keeps draining, the market's skepticism will look justified.