Why 800G and 400G NVIDIA Optics Have 40+ Week Lead Times in 2026 (And What to Do About It)

Short answer: genuine NVIDIA OEM 800G optics are running 40+ week lead times in 2026, and 400G isn't far behind at 16–24 weeks, according to industry sourcing analysts. The cause isn't a generic chip shortage — it's a narrow bottleneck in EML laser production that NVIDIA itself just moved to lock up $4 billion of, in strategic deals with Lumentum and Coherent announced March 2, 2026. Compatible modules like the ones we stock aren't magically exempt from the same underlying component crunch, but they aren't sitting in the same allocation queue as hyperscaler OEM orders either — which is the practical reason our restock lead time on the MMA4Z00-NS400 and MMA4Z00-NS pair currently runs 7–10 days, not 40 weeks.

Here's what's actually driving this, with sourcing, so you can plan a fabric build around it instead of getting surprised by it.

The Shortage, in Numbers

Figures reflect industry reporting as of August 2026 and will move as new capacity comes online — treat the ranges, not the exact week counts, as the durable takeaway.

Why This Is Happening: It's Lasers, Not Silicon

The instinct is to file this under "chip shortage." That's not quite right. The 2026 constraint is narrower and more specific: it's a bottleneck in EML (electro-absorption modulated) lasers — the component that encodes data onto the optical signal at the speeds 800G and 1.6T modules require. Fewer than five companies worldwide manufacture them at the volumes AI data centers now need: Lumentum, Coherent, Mitsubishi, Sumitomo, and Broadcom. One step further upstream, indium phosphide (InP) — the substrate those lasers are built on — is even more concentrated, with two to three suppliers controlling roughly 75% of global capacity.

On the assembly side, Chinese module makers Innolight and Eoptolink reportedly account for around 60% of NVIDIA's 800G module supply, and they're currently facing export restrictions on some of the advanced manufacturing equipment needed to expand that capacity further. None of these bottlenecks resolve with a fab ramp-up on a normal semiconductor timeline — optical fiber and laser capacity expansions run on 18–24 month cycles, and most forecasts don't expect meaningful relief before late 2027.

NVIDIA's Own Move Confirms the Problem

The clearest signal that this is real, and not just distributor talk, is what NVIDIA did about it. On March 2, 2026, NVIDIA announced strategic partnerships with both Lumentum and Coherent — $2 billion invested in each, alongside multiyear, nonexclusive purchase commitments described by Coherent as multibillion-dollar, to secure future EML laser and optical networking capacity for its own AI infrastructure roadmap.

That's not a company hedging against a hypothetical. That's the largest buyer of AI optics in the world spending $4 billion to make sure its own supply doesn't run out — which by definition means everyone else buying from the same small pool of laser manufacturers is competing for what's left. Multiple industry analysts covering the deal have noted it effectively pushes non-NVIDIA-aligned lead times further out, since Lumentum and Coherent's priority capacity is now spoken for.

Why Compatible Modules Aren't Stuck in the Same Queue

To be direct about it: a compatible MMA4Z00-NS or MMA4Z00-NS400 module is not immune to the physics of the EML laser shortage. The optics inside still come from a supply chain touched by the same upstream constraints as everyone else's.

What's different is allocation. NVIDIA's own factory-branded stock gets distributed through an authorized tier that prioritizes the largest, longest-standing hyperscaler and enterprise contracts first — that's how allocation always works when demand outstrips supply, and it's a large part of why the gap between OEM and compatible pricing has widened alongside the gap in lead time. Compatible-module suppliers work from ODM finished-goods inventory and smaller, faster-turning stocking cycles rather than build-to-order allocations tied to twelve-figure hyperscaler contracts. That's the honest reason a mid-size buyer sourcing a handful of ports sees a 7–10 day restock on a compatible module and a 40+ week quote on the equivalent OEM part — not a different laser, a different place in line.

We stock and test both halves of this specific pair against ConnectX-7 and Quantum-2 directly:

• MMA4Z00-NS400 flat-top (RHS) — $579, in stock, 7–10 day restock lead time on additional units
• MMA4Z00-NS finned-top (IHS) — $829, in stock, 7–10 day restock lead time on additional units

What to Actually Do If You're Speccing a Fabric Right Now

1. Quote both OEM and compatible before you commit a timeline. If a project deadline depends on optics arriving inside 90 days, get the OEM lead time in writing before you plan around it — don't assume list-price availability matches list price.
2. Order the pair at the correct ratio, not port-for-port. One MMA4Z00-NS switch-side module serves two MMA4Z00-NS400 server-side modules. Ordering 1:1 overstates the switch-side count by double, which matters more than usual when restocking takes over a week either way.
3. Don't wait for the shortage to "resolve" before locking in a supplier relationship. Forecasts point to late 2027 at the earliest for meaningful relief. A build that's ready this quarter shouldn't sit on a shelf waiting for a laser market that isn't turning around on that timeline.
4. Verify coding and warranty regardless of which path you choose. A tight market is exactly when corners get cut on testing. See our compatibility testing before you buy from anyone, us included.

FAQ

Is the 2026 transceiver shortage a general chip shortage?
No. It's a narrow, structural constraint centered on EML lasers and the indium phosphide they're built on — components made by a handful of companies worldwide — driven by AI data center demand specifically, not a broad semiconductor supply crisis. Legacy 10G/25G optics are largely unaffected.

Will NVIDIA OEM lead times improve before 2027?
Industry forecasts generally point to late 2027 at the earliest for 800G relief, and further out for 1.6T. New fiber and laser capacity takes 18–24 months to come online, and NVIDIA's own March 2026 investment in Lumentum and Coherent priority capacity adds further pressure on non-NVIDIA-aligned buyers in the meantime.

Does buying compatible instead of OEM avoid the shortage?
Not entirely — the underlying laser components are shared across the industry. What changes is allocation: compatible suppliers aren't competing in the same build-to-order queue as the largest OEM contracts, which is why lead times on compatible stock can run days instead of months.

Do I need the 400G or 800G module?
MMA4Z00-NS400 (flat-top) is the server/NIC-side part for ConnectX-7, ConnectX-8, BlueField-3, and DGX ports. MMA4Z00-NS (finned-top) is the switch-side part for air-cooled Quantum-2. One switch-side module serves two server-side modules. Flat-top or finned-top? covers the full decision.

What's covered under warranty?
Resilient Tec backs both modules with a lifetime warranty, reachable directly — not routed through a manufacturer ticket queue that's presumably just as backed up as everything else right now.

Resilient Tec stocks and tests the MMA4Z00-NS400 and MMA4Z00-NS against ConnectX-7 and Quantum-2 specifically, ships from Knoxville, Tennessee, and backs both with a lifetime warranty.

Ready to Buy?

• 400G OSFP SR4 Flat-Top (RHS) — MMA4Z00-NS400 / 980-9I51S-00NS00 compatible — $579, in stock
• 800G OSFP SR8 Finned-Top (IHS) — MMA4Z00-NS compatible — $829, in stock
• NVIDIA OSFP part-number crosswalk — confirm your exact part before the PO goes out

Sizing a fabric before the lead-time math gets worse? Request a quote and tell us the port count — we'll price both sides of the link together at the correct ratio.

You Might Also Need

• Is a third-party MMA4Z00-NS / NS400 safe? — what to verify before you buy compatible optics for this platform
• Flat-top vs finned-top OSFP — the physical/thermal decision behind ordering the right module
• OSFP DAC cables — when the run is short enough that copper sidesteps the optics market entirely
• Optical transceivers guide — 1G SFP through 800G OSFP, across platforms
• Network switches — Arista, Cisco, HPE/Aruba and Mellanox, new and tested pre-owned