WillItSurvive

Contract manufacturer building AI servers and racks for AMD and hyperscalers.

Losing its protection
Assessed 20 July 2026 · Evidence as of 20 July 2026
The structural protection is eroding.
On the recordSince 20 July 2026 · Checkable from around July 27, 2026 · Every call on the record Checkpoint: Q3 FY2026 earnings report
The call

Sanmina is riding a real AI hardware wave with AMD's business, but it is running the same race as Celestica and Foxconn on thin margins with no lock stopping either from closing the gap.

Sanmina is a manufacturer that builds server components and equipment for companies like AMD and hyperscalers (the giant cloud companies). It is reasonably hard to replace because AMD treats it as a preferred partner and customers have existing relationships with it, but this advantage is not unique, since competitors like Celestica and Jabil are doing the exact same work and Celestica is actually growing faster at it. The real risk is that Sanmina's profits are razor-thin, so if the artificial intelligence data-center spending slows down or if Foxconn (a much larger manufacturer) decides those margins are worth entering, Sanmina could get squeezed hard.

Why
01
AI pressure: None apparent. AI infrastructure demand is feeding Sanmina's growth right now, not threatening it; the risk is the capex cycle turning, not AI itself replacing this work. Could AI do the core job? Unlikely.
02
Bigger players: Celestica and Foxconn. Celestica runs the same AI rack playbook and is growing faster; Foxconn is 5 to 10 times bigger and can step in once margins look attractive. Could a giant copy it? Yes, easily.
03
What protects it. Qualified-supplier status plus AMD preferred NPI role and 42Q platform give it a head start, but nothing stops AMD from adding a second partner.
What would change this call
Celestica reports AI-infrastructure revenue growth versus Sanmina over the next 2-3 quarters. Shows whether Sanmina's AMD-ZT Systems head start is durable or being overtaken.
AMD names a second preferred NPI manufacturing partner besides Sanmina. Would show Sanmina's AMD relationship is diluting rather than exclusive.
Hyperscaler capex guidance flattens or is cut in upcoming earnings from Microsoft, Meta, or Amazon. Would test whether Sanmina's thin margins can survive a slowdown in AI infrastructure spending.
Key numbers
  • FY2025 revenue $8.1B, operating margin 4.4% (2025-09-30)
  • Q2 FY26 revenue $4.01B, up 102.3% YoY (2026-03-28)
  • FY2027 revenue target $16B+ (2026-04-27)
Named risks
  • AI infrastructure capex slowdown would directly hit FY2027 $16B+ revenue target dependent on hyperscaler spend. Severity: high
  • ZT Systems integration execution failure could derail margin accretion and growth guidance. Severity: medium
  • Thin ~4.4% operating margins leave no cushion if AI demand cycle peaks or slows. Severity: high
  • Tariff and trade policy uncertainty (Section 301/232) could raise costs given Mexico/Asia manufacturing footprint. Severity: medium
  • AMD could dilute Sanmina's preferred-partner status by adding a second NPI manufacturing partner. Severity: medium
Rivals, as read in July 2026
RivalDirectionEvidence
Celesticalosing toCited as 'clearest expression' of AI rack-integrator thesis, growing faster than Sanmina as of 2025-12-09
Foxconncontesting5-10x larger incumbent EMS giant that can enter AI rack integration when margins justify, per 2026 brief
JabilcontestingRacing into identical AI/data-center rack integration pivot as Sanmina, no structural block cited (2025-12-09)
FlexcontestingAlso pursuing same AI rack-integration opportunity as Sanmina, per December 2025 analysis
Questions people ask
Is Sanmina profitable in 2026?
Yes. Sanmina posted FY2025 net income of ~$246M on $8.1B revenue (ended 2025-09-30), and Q2 FY26 (ended 2026-03-28) revenue hit $4.01B, up 102.3% YoY, with non-GAAP EPS of $3.16 beating the $2.42 consensus.
Could AI replace Sanmina's manufacturing business?
No, physical rack assembly, liquid-cooling engineering, and certifications like AS9100D can't be automated away. AI infrastructure demand is actually a tailwind: AI/cloud made up 61.5% of Q1 FY26 revenue (as of 2026-01-31), up from 36.7% a year earlier.
Who is Sanmina's biggest competitor in AI infrastructure?
Celestica is cited (as of 2025-12-09) as the 'clearest expression of the AI rack integrator thesis,' growing faster than Sanmina on the identical hyperscaler pivot. Jabil, Flex, and much-larger Foxconn are also chasing the same AI rack-integration opportunity.
Why did Sanmina buy ZT Systems from AMD?
Sanmina completed the $3 billion acquisition of AMD's ZT Systems data-center manufacturing business on 2025-10-27, gaining rack-scale AI/liquid-cooling manufacturing and becoming AMD's preferred new-product-introduction partner for cloud rack and cluster-scale AI systems.
What is Sanmina's revenue guidance for fiscal 2026?
As of 2026-04-27, Sanmina guided full FY2026 revenue of $13.7 to $14.3 billion, with core Sanmina growing high single digits and ZT Systems contributing $5 to $6 billion annualized; management also targets $16 billion-plus revenue for FY2027.
Are Sanmina insiders selling stock?
Yes. CEO Jure Sola sold approximately $27 million in Sanmina stock in May 2026, according to a 2026-07-20 MarketBeat report, adding to investor uncertainty even as the stock has traded in a volatile $95.49 to $288.68 52-week range.
Timeline
  • 2026-04-27Q2 FY26 revenue hits $4.01B, up 102.3% YoY; board authorizes $600M buyback
  • 2026-03-1042Q platform reported linking 70 factories, 15 countries, ~35,000 machines
  • 2025-12-31Q1 FY26 revenue $3.19B, up 59% YoY
  • 2025-10-27Sanmina completes $3B acquisition of AMD's ZT Systems manufacturing unit
  • 2025-09-30FY2025 closes with $8.1B revenue, ~$355M operating income, ~$246M net income
  • 2025-05-19AMD announces agreement to divest ZT Systems to Sanmina
  • 2020-08Jure Sola returns as Chairman and CEO
  • 2012Company reverts name from Sanmina-SCI back to Sanmina
  • 2001Sanmina merges with SCI Systems, forming Sanmina-SCI
  • 1980Sanmina founded by Jure Sola and Milan Mandaric
Revision history
20 July 2026
First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources