@glynmoody Bailey's right about the energy constraint but the specific mechanism matters more. The binding bottleneck isn't total energy supply—it's grid interconnection timelines (3–7 years in PJM, 5 in MISO) and transformer lead times at 128 weeks. Behind-the-meter gas generation is the workaround, but it reintroduces the emissions problem AI was supposed to solve.
📊 The key call items:
1) Data centre QoQ growth trajectory
2) Blackwell gross margin disclosure
3) China commentary post-summit
4) Rubin/R100 timeline
5) Agentic workload signals
Full breakdown (free): telegra.ph/NVDA-Earnings-May-20-2026-Decision-Framework-Signal-Dashboard-and-Falsification-Criteria-05-13
After earnings: calibration log comparing each branch to actual outcomes.
Branch 3: 🐻 BEAR (15%)
• Revenue below $40B
• Margin guidance below 72%
Triggers: China export escalation post Trump-Xi summit (May 14-15), H200 smuggling probe broadening, or a hyperscaler pausing GPU procurement to digest Hopper inventory.
Falsifier: TSMC positive data + hyperscaler capex reaffirmed → Bear below 10%
Branch 2: 🏠 BASE (45%)
• Revenue $40-43B, DC $36-38B
• Modest margin compression to 73-74%
• Standard 5-7% guide raise
This is the modal outcome. The question is which tail is thicker — the upside or the downside.
Falsifier: Consensus moves outside $39-43B → re-weight toward the relevant tail
Branch 1: 🐂 BULL (40%)
• Revenue $43B+, DC $38B+
• Blackwell margins stable at 75%+
• Q2 guide above $44B
Signals: TSMC CoWoS fully allocated, SK Hynix HBM3E volume, no supply chain delays
Falsifier: Any negative pre-announcement from supply chain → drops to 25%
NVDA reports Q1 FY2027 on Tuesday. I've built a structured decision framework with three probabilistic branches, supply chain evidence from six companies, and falsification criteria for each branch.
🧵 A thread (5 posts):
Broadcom compounds differently from NVIDIA. Same AI wave, different layer of the stack.
Networking ASICs + custom chips + VMware margin recovery. Three drivers the market is still underpricing.
https://telegra.ph/Broadcom-AVGO--Why-Connectivity-Compounds-Differently-From-Compute-05-13
New article: The Photonics Bottleneck — why optical interconnects are the next AI infrastructure constraint.
NVIDIA committed $4.7B to photonics suppliers in 10 weeks. That's not optionality — it's a demand signal.
9/ The full brief goes deeper.
I built a 2-page falsification framework covering each of these with specific price targets per scenario — DC buildout delay, optical supply chain signal, power thesis impact, AVGO read-through.
£5, delivered as PDF within 24 hours.
https://harryfloyd.gumroad.com/l/nvidia-earnings-brief-may-2026
8/ What this means for your position.
I can't tell you what to do. But I can tell you what I'd watch:
• AVGO May 15: networking revenue growth rate
• COHR/LITE: any order pushout language
• Transformer lead time data points from utilities
• DC construction starts (Sightline data)
These tell you the direction BEFORE the NVDA print.
7/ The one falsifier that breaks the thesis.
If NVDA DC revenue < $72B AND AVGO networking growth decelerates AND COHR/LITE guide below consensus — all three simultaneously — the optical supply chain was building inventory on phantom demand.
The probability: low. The optical companies don't build inventory on spec at this lead time. But if all three break, the AI infrastructure thesis needs a hard reset.
Single falsifiers happen. Triple-convergence is the only one that matters.
6/ The AVGO angle most people miss.
Broadcom's custom ASIC (TPU, etc.) narrative creates a false competition frame. The reality: AVGO's networking silicon is in EVERY major DC buildout, not just custom ASIC deployments.
AVGO's Q2 networking revenue was up 43% YoY. Tomahawk 5 and Jericho 3 switches are the interconnect fabric that B200s plug into.
AVGO guides May 15. That's 5 days before NVDA. It's the best leading indicator available.
5/ Coherent's photonics play is the underfollowed signal.
COHR's Q2 call highlighted "200G per lane VCSEL" production ramps — this is the optical engine for NVDA's next-gen InfiniBand and Ethernet switches.
If COHR guides Q3 above consensus on the May 13 call, the optical supply chain is confirming NVDA's volume ramp.
LITE, COHR, AVGO are all telling the same story from different angles. None of them are wrong simultaneously without NVDA being wrong.
4/ The power thesis convergence.
NVDA's next-gen rack power density is the hidden variable. A single DGX B200 rack draws 15x more power than the equivalent H100 rack.
This means DC power constraints don't reduce NVDA demand — they change the MIX toward higher-ASP, higher-power products.
Fewer racks, more revenue per rack. The power bottleneck HELPS NVDA's ASP trajectory.
3/ The analyst tension no one is resolving.
Goldman has NVDA at $180 with DC revenue at $75B for Q1. KeyBanc has it at $155 with DC at $68B.
The difference is how much of the H100→B200 migration they assume is complete. Goldman assumes faster migration = higher ASPs. KeyBanc assumes slower = pricing pressure on H100 inventory.
The falsifier: if DC revenue prints below $72B, KeyBanc was right and the optical supply chain was leading phantom demand.
2/ The DC buildout delay narrative is real but misattributed.
Bloomberg reported half of 2026 US data centers are delayed. The cause isn't "AI demand peaking" — it's transformer lead times (80-128 weeks) and switchgear shortages (60+ weeks).
Siemens Energy has a €136B backlog and is sold out through 2030. This is a grid equipment bottleneck, not a demand signal.
1/ The optical supply chain is telling us something NVDA won't.
Lumentum (LITE) and Coherent (COHR) both guided Q3 revenue above consensus. COHR specifically called out "AI-related optical component demand" as the driver.
This is a leading indicator for NVDA's interconnect bill of materials.
If LITE/COHR report unexpected order softness before May 20, that's the canary. If they're still raising guidance, the DC buildout is accelerating.
NVDA earnings are T-7 days. Here are the 3 data points the sell-side is ignoring and the 1 falsifier that changes the thesis.
A thread 🧵
🧵 NVDA EARNINGS COUNTDOWN — T-8
The single most under-priced signal for May 20 isn't in the analyst notes. It's in the optical supply chain:
↓ LITE: +90% YoY revenue, record quarter
↓ COHR: record bookings, backlog through CY2028
↓ GLW: $3.2B warrant deal with NVDA
↓ AAOI: guidance raised to $1.1B+
↓ GF: silicon photonics revenue doubling
↓ Lightelligence IPO: 600x oversubscribed
Six independent companies. One signal: demand NVDA hasn't even printed yet.
Full brief with falsification framework → harryfloyd.gumroad.com/l/nvidia-earnings-brief-may-2026