Sign in

You're signed outSign in or to get full access.

GT

GSI TECHNOLOGY INC (GSIT)·Q2 2026 Earnings Summary

Executive Summary

  • Q2 FY2026 revenue was $6.44M, up 3% sequentially and 41.6–42% YoY; gross margin fell to 54.8% from 58.1% in Q1 on product mix; diluted EPS was $(0.11) versus $(0.08) in Q1 and $(0.21) YoY .
  • Management guided Q3 FY2026 net revenues to $6.0–$6.8M and gross margin to 54–56%, implying a similar revenue range but a lower GM midpoint versus Q2 guidance issued last quarter (56–58%) .
  • Strategically, GSIT closed a $50M registered direct financing to accelerate Gemini‑II software and commence the Plato chip design; cash ended Q2 at $25.3M before the financing’s October close, with equity at $38.6M .
  • Cornell research validated Gemini‑I APU’s GPU‑class performance with >98% lower energy on RAG workloads, strengthening the edge‑AI positioning; an SDA SBIR extension of ~$752K supports radiation tolerance work on Gemini‑II .
  • Near‑term stock catalysts include continued defense POCs (SAR, multimodal LLM edge benchmarks by year‑end), Q3 results against tightened GM guidance, and progress on Plato IP acquisition and partner engagement .

What Went Well and What Went Wrong

What Went Well

  • Strong top‑line momentum: net revenue +3% QoQ and +41.6–42% YoY as SRAM demand improved; SigmaQuad mix rose YoY (50.1% vs. 38.6%) .
  • Third‑party validation: “Cornell…validated our Gemini‑I APU's ability to match NVIDIA's A6000 GPU performance…while consuming over 98% less energy,” reinforcing GSIT’s compute‑in‑memory architecture advantage .
  • Capital secured for roadmap: “closed a $50 million registered direct offering to fund continued development of our APU product line…support…Gemini‑II software and…Plato chip design” .

What Went Wrong

  • Margin compression vs. Q1: GM fell to 54.8% from 58.1% due to product mix, pressuring profitability (operating loss widened to $(3.19)M; net loss $(3.19)M) .
  • Opex increased QoQ: total operating expenses rose to $6.7M (R&D $3.8M; SG&A $3.0M), reflecting higher stock‑based comp and variable SBIR offsets vs. Q1 .
  • Customer concentration and Nokia softness: Cadence represented 21.6% of revenue; Nokia fell to 3.1% (from 17.8% YoY), raising sustainability questions for mix and repeatability .

Financial Results

MetricQ2 FY2025Q4 FY2025Q1 FY2026Q2 FY2026
Revenue ($USD Millions)$4.55 $5.88 $6.28 $6.44
Gross Margin (%)38.6% 56.1% 58.1% 54.8%
Operating Loss ($USD Millions)$(5.58) $(2.28) $(2.18) $(3.19)
Net Loss ($USD Millions)$(5.46) $(2.23) $(2.22) $(3.19)
Diluted EPS ($USD)$(0.21) $(0.09) $(0.08) $(0.11)

Segment/customer and mix detail:

ItemQ2 FY2025Q1 FY2026Q2 FY2026
KYEC Sales ($000, % of Rev)$650; 14.3% $267; 4.3% $802; 12.5%
Nokia Sales ($000, % of Rev)$812; 17.8% $536; 8.5% $200; 3.1%
Cadence Design Systems ($000, % of Rev)$0; 0% $1,500; 23.9% $1,400; 21.6%
Military/Defense Shipments (% mix)40.2% 19.1% 28.9%
SigmaQuad Shipments (% mix)38.6% 62.5% 50.1%

Operating metrics and KPIs:

MetricQ1 FY2026Q2 FY2026
Total Operating Expenses ($USD Millions)$5.83 $6.72
R&D Expense ($USD Millions)$3.10 $3.77
SG&A Expense ($USD Millions)$2.73 $2.95
Pre‑Tax SBC ($USD Thousands)$341 $856
Cash & Equivalents ($USD Millions)$22.7 (6/30/25) $25.3 (9/30/25)
Working Capital ($USD Millions)$25.7 (6/30/25) $26.8 (9/30/25)
Stockholders’ Equity ($USD Millions)$37.4 (6/30/25) $38.6 (9/30/25)

Estimates comparison (S&P Global):

MetricQ2 FY2026 ActualQ2 FY2026 Consensus (S&P Global)
Revenue ($USD Millions)$6.44 N/A*
Diluted EPS ($USD)$(0.11) N/A*
*Values retrieved from S&P Global.

Guidance Changes

MetricPeriodPrevious GuidanceCurrent GuidanceChange
Revenue ($USD Millions)Q2 FY2026$5.9–$6.7 Actual: $6.44 Achieved toward high end
Gross Margin (%)Q2 FY202656–58 Actual: 54.8 Below guided range
Revenue ($USD Millions)Q3 FY2026—$6.0–$6.8 New; similar range to Q2
Gross Margin (%)Q3 FY2026—54–56 New; midpoint lower vs Q2 guide

Earnings Call Themes & Trends

TopicQ4 FY2025 (Q‑2)Q1 FY2026 (Q‑1)Q2 FY2026 (Current)Trend
AI/APU validationAnticipated Gemini‑II readiness; Plato enhancements (camera interface) Gemini‑II second silicon production‑ready Cornell paper validates GPU‑class perf at 98% less energy Strengthening validation
Defense POCs/SARAFRL server shipped; SDA card pending; rad‑hard SRAM order Delivered Leda‑2 board to offshore defense contractor SAR low‑power (~15W) POC approved; SDA SBIR extension ~$751–$752K Pipeline building
Multimodal LLM at edgeHighlighted Plato edge applications Developing multimodal LLM; benchmarks by fall 2025 YOLO + LLM time‑to‑first‑token edge demo; benchmarks by YE 2025 Execution progressing
Financing/CapitalNo major financing disclosedCash up via ATM; Q2 guide set Closed $50M registered direct offering De‑risked funding
Data center vs. edge focusConsidering Gemini‑3 data‑center concept —Staying focused on edge; data center requires different partner/funding Edge emphasis sustained
Customer mixKYEC/Nokia/Defense variability Cadence $1.5M (23.9%); Defense 19.1% Cadence $1.4M (21.6%); Defense 28.9% Concentration persists

Management Commentary

  • “The research paper validated our Gemini‑I Associative Processing Unit's ability to match NVIDIA's A6000 GPU performance…while consuming over 98% less energy…With 8× memory and 10× performance…Gemini‑II is poised to deliver even greater processing power at dramatically lower energy” — Lee‑Lean Shu, CEO .
  • “We closed a $50 million registered direct offering to fund continued development of our APU product line…support…the progression of Gemini‑II software and the initiation of the Plato chip design” — Lee‑Lean Shu .
  • “Gross margin…was primarily due to a change in the product mix” — Douglas Schirle, CFO .
  • “Gemini‑II has been approved for prototyping…for SAR…Our solution delivers the required performance while maintaining ~15 watts” — Didier Lasserre, VP Sales .

Q&A Highlights

  • Strategic focus: Management reiterated prioritization of edge applications over data centers given resource constraints; data‑center efforts (Gemini‑3) would require strategic partner/funding; licensing or co‑development remains possible but not active now .
  • Validation plans: Discussed providing Gemini‑II boards to Cornell and other researchers for third‑party validation following Gemini‑I paper .
  • Commercialization milestones: Pilot shipments underway with more in CY2026; POCs could drive more substantial revenue in H2 CY2026; Plato tape‑out targeted for early CY2027; ~$15–$17M in fixed IP/mask costs, with remaining funds split between Gemini‑II and Plato engineering .
  • TAM/pricing/margins: Edge AI drone TAM sizable per Needham; boards priced in the low thousands to $10K; chips ~$1K+; target gross margins 60–80% by market/use case (above current corporate GM) .
  • Break‑even framing: CFO suggested 65–70% gross margin could support breakeven depending on hiring/plans; specifics TBD .

Estimates Context

  • S&P Global consensus for Q2 FY2026 EPS and revenue appeared unavailable for GSIT this quarter; actuals were Revenue $6.44M and diluted EPS $(0.11) .
  • Given Q3 guidance implies similar revenue range but a lower GM midpoint, Street models (where covered) may need to adjust gross margin assumptions downward for near‑term quarters while keeping revenue broadly stable.*
    *Values retrieved from S&P Global.

Key Takeaways for Investors

  • Revenue trajectory is improving with two consecutive quarters above $6M and ~42% YoY growth, but near‑term margins are sensitive to mix; expect GM volatility around the 54–56% guided band .
  • The $50M financing materially de‑risks execution of Gemini‑II software stack and Plato IP acquisition/tape‑out, extending runway and enabling parallel hardware/software development .
  • Third‑party validation (Cornell) and defense POCs (SAR, multimodal LLM) bolster the edge‑AI thesis; watch for year‑end benchmark disclosures and additional defense program traction as potential catalysts .
  • Customer concentration (Cadence >20% of revenue) and Nokia softness warrant monitoring; diversification and scaling defense/commercial customers are key to sustaining growth and margin mix .
  • Short‑term trading: Q3 gross margin guide is lower vs prior; print in line on revenue with stable mix could be neutral; upside hinges on favorable product/customer mix lifting GM toward high end of the 54–56% range .
  • Medium‑term thesis: If GSIT converts POCs to production and executes Plato on timeline, structural margins could expand toward targeted 60–80% in select board/software bundles, shifting earnings power materially .
  • Risk management: Execution on compiler/SDK/tooling and expanding an ecosystem are critical; management acknowledged staged tooling development with partners/customers in progress .