GSIT

GSIT is two businesses stapled together: a shrinking legacy Very Fast SRAM operation (networking/telecom, which management explicitly says 'has declined during the past several years and is expected to continue to decline') funding a pre-revenue bet on in-place associative computing (Gemini-II, the Plato pipeline, APU similarity/vector search). FY2026 revenue rose 22.4% to $25.1M and gross margin expanded to 54.5%, but that only claws back a fraction of the multi-year slide from $33.4M (FY2022) to $20.5M (FY2025) — the top line is volatile, customer-concentrated (Nokia, KYEC, Cadence), and still below where it was four years ago. Critically, the actual growth engine isn't earning yet: the 10-K states the APU products 'have not resulted in material revenues to date.' You are paying 10.4x sales for a company whose thesis rests on products that don't yet sell.

The loss profile is the tell. Operating margin is -69.6%, net loss -$13.2M, and operating cash flow -$15.9M — R&D alone ($19.9M) is 79% of total revenue, so the legacy business cannot self-fund the APU pivot. Management said so directly: it is 'using the revenue generated by the sales of high-speed synchronous SRAM products to finance the development' of the new lines, and warns it 'may need additional funding to complete the commercialization and development of Gemini-II and Plato.' That funding gap has already been filled by shareholders, not by the business: a May–August 2025 ATM raise ($14.3M net at ~$3.29), plus a $50M registered direct offering in October 2025 at $10.00. Together those drove shares outstanding up 37.2% and equity up 189.7% in a single year. The pristine balance sheet — $67.2M cash, no debt, liabilities/equity 0.20x — is real and buys multiple years of runway, but it was purchased with heavy, ongoing dilution, and the current $6.82 price sits well below the $10 institutional investors paid nine months ago.