Start here if you're new
what it is
GSI makes specialty memory chips and a search chip for networking, industrial, medical, aerospace, and military gear.
how it gets paid
Last year Gsi Technology made $21M in revenue. SyncBurst SRAM was the main engine at $8.4M, or 40% of sales.
why growth slowed
Revenue fell 5.7% last year. 55.2% gross margin is the tell. The product still works.
what just happened
A single quarter scales to roughly ~$5M of revenue (~one-fourth of ~$21M for the year); the latest print still showed a loss of about $0.28 per share.
At a glance
C++ balance sheet — some cracks in the foundation
40/100 earnings predictability — expect surprises
~-$0.28 latest-quarter EPS
~$21M annual revenue (FY bridge)
-52.8% operating margin — deep operating loss
xvary composite: 49/100 — below average
What they do
GSI makes specialty memory chips and a search chip for networking, industrial, medical, aerospace, and military gear.
GSI has 121 employees and $21M of annual revenue. That is a tiny team carrying a public chip business. You are buying niche memory parts and an APU for similarity search, not a commodity chip made by the ton. The contrast is the whole story: 55.2% gross margin (money left after making the chips) versus -52.8% operating margin (money left after running the company).
How they make money
$21M
annual revenue · their business grew -5.7% last year
SyncBurst SRAM
$8.4M
No Bus Turnaround and Sigma SRAM
$5.0M
Associative processing unit products
$3.6M
Low Latency DRAM
$2.1M
Asynchronous SRAM
$1.9M
The products that matter
legacy memory products
SigmaQuad SRAM
41.7% of shipments last quarter
It represented 41.7% of product shipments last quarter, down from 50.1% the prior quarter. That's still a core revenue source, but not one growing fast enough to carry the story.
legacy core
ai inference accelerator
Gemini APU
$0.5M in APU & other revenue
This is the strategic bet. It targets a $100B AI inference market, but the current revenue line for APU & other is just $0.5M. Right now, the opportunity is much larger than the business.
option value
Key numbers
$21M
annual revenue
You are looking at a $21M chip business that still lost money on operations.
-52.8%
operating margin
For every $1 of sales, overhead and R&D consumed more than half a dollar beyond gross profit — the business is loss-making at the operating line.
55.2%
gross margin (FY)
The chips themselves still make money before overhead gets paid.
1.6
beta
The stock tends to move about 60% more than the market.
Financial health
C++
strength
- balance sheet grade C++ — below average — limited financial resources
- risk rank 2 — safer than 80% of stocks
- price stability 5 / 100
- long-term debt $7M (2% of capital)
C++ — risk rank looks solid but balance sheet grade needs watching.
Total return vs. market
Return history isn't available for GSIT right now.
source: institutional data · return history unavailable
What just happened
missed estimates
GSI posted roughly ~$5M in quarterly revenue (~¼ of ~$21M FY) and still lost $0.28 a share.
Prior “$19M quarter” lines treated a longer window as one quarter. EPS was -$0.28. Gross margin was 55.2%, so the quarter was a spending problem, not a pricing problem.
~$5M
revenue (q)
-$0.28
eps (q)
55.2%
gross margin (q)
margin held
55.2% gross margin is the tell. The product still works, but the company spent too much to turn it into profit.
source: company earnings report, 2026
Get this snapshot in your inbox
This page, delivered free — plus weekly updates when the numbers change. plain english, no spam.
weekly updates
earnings alerts
plain english
no spam
What could go wrong
the #1 risk is Gemini and Plato failing to become a real revenue business before legacy memory keeps shrinking.
med
APU commercialization failure
The valuation leans on Gemini and Plato, but APU & other contributed just $0.5M of a $21M revenue base. That is not traction. That is an early-stage idea attached to a public stock.
If that revenue stays stuck near the current level, the 11.8x sales multiple has very little fundamental support.
med
Legacy memory keeps eroding
Legacy memory products still generate 97.6% of revenue, and that segment declined 5.7% last year. The company is trying to replace almost the entire business while the current one is already moving backward.
A few more quarters of decline would make the transition story more urgent and less funded.
med
Institutional support vanished fast
Institutional ownership dropped from 18.5% in September 2025 to 2.7% in December 2025. In small caps, that can change how the stock trades even before it changes how the business performs.
Lower professional ownership can mean thinner liquidity, sharper price swings, and less patience for missed milestones.
med
Foundry dependence limits the upside path
As a fabless chip company, GSI still depends on outside manufacturing capacity to scale any APU success. That is manageable when volumes are small. It gets harder when a real product launch needs wafers on time.
Even good customer interest would not matter much if production slips or unit economics arrive later than expected.
97.6% of revenue still comes from legacy memory, while the stock trades at 11.8x sales for what the AI business might become next.
source: institutional data · regulatory filings · risk analysis
Pay attention to
earnings
Q4 FY2026 revenue versus the $5.7M–$6.5M guide
The guide is the next reality check. If revenue lands near the low end again, the AI transition still is not showing up in the numbers.
revenue mix
APU & other revenue moving above the current $0.5M base
This is the cleanest scoreboard on the page. Until that line starts growing, most of the valuation is still future tense.
margin
Gross margin after the drop to 52.7%
Margins are still decent, but down is down. If the company is scaling into AI, you want better mix over time, not slippage.
product milestone
Plato tape-out and Gemini-II proof-of-concept progress
Tape-out is scheduled for early 2027, and management has pointed to a government-funded Gemini-II proof-of-concept. If either slips, the market has less reason to pay an AI premium now.
Analyst rankings
earnings predictability
40 / 100
in human-speak, analysts do not view this as a smooth numbers story.
balance sheet grade
C++
Enough financial support for now, but not enough scale to hide execution mistakes.
beta
1.6
When the market moves, this stock has tended to move more. Not a bunker stock.
source: institutional data
Institutional activity
institutional ownership data for GSIT is being compiled.
source: institutional data
Price targets
3-5 year target range
n/a
n/a
$9
current price
n/a
target midpoint · n/a from current
Want the deeper analysis?
The full deep dive: dcf model, scenario analysis, competitive moat breakdown, and quarterly tracking — everything on this page, taken further.
see plans from $5/moThe deep dive