Training backend = rail-optimized collectives. Inference backend = lossless RDMA for P/D disaggregation, KV cache, isolation. We run OCP Inference Fabric RA as our backend fabric: DS5000 + SONiC + @githedgehog declarative GitOps. @Celestica_Inc opencompute.org/documents/op…

OCP NVMe SMART telemetry is richer than most GPU telemetry — we just don't use it. Most GPU clouds monitor GPU temp, GPU util, and call it observability. The OCP Datacenter NVMe spec exposes 30+ telemetry dimensions per drive that most fleets never collect: NAND wear curves, ECC escalation rates, recoverable error histograms, thermal throttling counters, grown defect list growth, power-state transition latency. We built a fleet-wide predictive failure engine on top of that surface. Every drive in 3 sites scored for degradation trends. Replace / Watch / Monitor — 3 weeks before failure, not after. For clarity: base NVMe SMART / Health Log is foundational. % used, media & data integrity errors, composite temp, unsafe shutdowns, error log depth. If you're not alerting per-drive per-hour on it, you're flying blind. We do. OCP doesn't replace SMART — it extends it. Endurance group histograms, physical media error bucketization, thermal history, PLP event counters, grown defect trending. SMART = what *is* wrong. OCP = where it's *going*. That's where 10+ years of OCP + NVMe storage work matters. We run both layers together: SMART for immediate health, OCP telemetry for slope. That's how you go from reactive RMA to scheduled replacement. Storage isn't secondary to GPU reliability — it *is* GPU reliability. 🌾

OCP survey of 19 neocloud operators: top pain points are RoCE tuning, firmware management, and BMC/BIOS inconsistency. Translation: every neocloud under 500 racks is independently re-discovering the same failure modes and filing the same tickets to the same 3 OEMs. This is the neocloud ops tax. Not CapEx. Not power. It's the $2M/yr in hidden engineering time spent building bespoke workarounds for Redfish endpoints that work on Dell and break everywhere else. For NIC firmware that needs per-rail tuning. For SONiC versions that handle ECN differently. FarmGPU + Hedgehog + Celestica built a reference fabric specifically to kill this tax: Celestica DS5000 (51.2T) + SONiC + Hedgehog K8s-native control. 36 composition variants, declarative, vendor-neutral. Contributed to OCP so other neoclouds don't have to reinvent it. The whole thesis: write down what neoclouds actually need -> get OEMs to ship it through OCP -> open-source the operator tooling that makes the spec real. Neoclouds don't need another GPU SKU. They need to stop paying that ops tax. blog.farmgpu.com/ocp-scaling… 🌾🚜

Meta contracted 5+ GW of data center capacity in H1 2026 alone — and nobody's building enough. Now they're doing "SpaceX-style" bulk compute leasing. New market segment, much higher revenue per megawatt than traditional cloud. The compute shortage isn't ending. Dylan Patel projects OpenAI + Anthropic could demand 100+ GW by 2030. FarmGPU's capacity is the asset. The cloud platform is just the interface. newsletter.semianalysis.com/…

SK Hynix just removed the price cap from its long-term memory contracts. On its own, a contract structure change. In context, a signal about who holds pricing power in the AI stack. For years, memory LTAs had ceilings — when spot prices spiked, contract buyers were protected. Micron still does this. SK Hynix is now saying: no ceiling. If the market goes higher, you pay more. With DDR5 retail having roughly tripled since October, there's a lot of upside to be exposed to. This splits the memory market into two regimes. Micron customers get predictability at very high Q2'26 levels. SK Hynix customers get uncapped exposure. And SK Hynix is the dominant HBM3E supplier for NVIDIA — so the downstream repricing runs through SSD controllers (which need DDR5 buffers), through eSSD procurement, through GPU cloud TCO, all the way to inference cost per token. The deeper signal: SK Hynix believes this shortage lasts long enough that uncapped contracts beat capped ones. They're willing to risk customer relationships on it. When the dominant HBM supplier for the AI accelerator market decides pricing power is theirs to keep, every layer downstream reprices. The memory shortage isn't a cycle — it's becoming the architecture of AI cost.

OCP just released the AI Inference Fabric Reference Architecture — the companion to the Training Fabric RA we co-developed with Hedgehog and Celestica. It's built on the exact stack FarmGPU runs in production: Celestica DS5000 (51.2T) + SONiC + Hedgehog K8s-native control plane. Inference ≠ Training. Different fabric demands: → North-south throughput for model serving → Multi-tenant isolation → Gateway services for ingress/egress → Lossless RDMA across hosts for distributed inference 64–1024 xPU clusters. Open, vendor-neutral, declarative. Production-ready blueprints instead of slideware. This is how neoclouds stop reinventing the network for every deployment. opencompute.org/documents/op… 🌾🚜

Shepherd: 51 AI agents watching your cluster so you don't have to. Every farm runs out of time, eyes, and bandwidth. Shepherd agents never do. 51+ autonomous agents across 8 domain crews — power, storage, network, security — operating on a 5-minute loop. CrewAI orchestrator. Fine-tuned triage model that classifies incidents before you wake up. No dashboards. No alerts you have to manually check. Just agents that fix things, escalate what matters, and keep your cluster alive at 3 AM. blog.farmgpu.com/b200-cluste… 🌾🚜

OCP just released the AI Inference Fabric Reference Architecture — the companion to the Training Fabric RA we co-developed with Hedgehog and Celestica. It's built on the exact stack FarmGPU runs in production: Celestica DS5000 (51.2T) + SONiC + Hedgehog K8s-native control plane. Inference ≠ Training. Different fabric demands: → North-south throughput for model serving → Multi-tenant isolation → Gateway services for ingress/egress → Lossless RDMA across hosts for distributed inference 64–1024 xPU clusters. Open, vendor-neutral, declarative. Production-ready blueprints instead of slideware. This is how neoclouds stop reinventing the network for every deployment. opencompute.org/documents/op… 🌾🚜

OCP just released the AI Inference Fabric Reference Architecture — the companion to the Training Fabric RA we co-developed with Hedgehog and Celestica. It's built on the exact stack FarmGPU runs in production: Celestica DS5000 (51.2T) + SONiC + Hedgehog K8s-native control plane. Inference ≠ Training. Different fabric demands: - North-south throughput for model serving - Multi-tenant isolation - Gateway services for ingress/egress - Lossless RDMA across hosts for distributed inference 64–1024 xPU clusters. Open, vendor-neutral, declarative. Production-ready blueprints instead of slideware. This is how neoclouds stop reinventing the network for every deployment. opencompute.org/documents/op…

Leopold Aschenbrenner’s Situational Awareness LP just dropped its Q1 2026 13F, and the move that stands out isn’t the longs — it’s the puts. The fund layered on multi-billion dollar bearish positions across the semiconductor stack: more than billion notional on SMH, over .5 billion on NVDA, roughly billion on AVGO, plus sizable books on AMD, TSM, ASML, MU, and fresh INTC puts while closing prior long calls. Last quarter they held basically zero puts. This is not protective hedging. This is a deliberate, sized short on the most crowded, highest-multiple corner of the AI hardware trade. The deeper context is Aschenbrenner’s core thesis from his 2024 essay “Situational Awareness: The Decade Ahead.” He puts AGI around 2027 and superintelligence shortly after, but argues the binding constraint is physical infrastructure, not intelligence. Trillions in capital expenditure will be required for clusters, power generation, chips, cooling, and data centers. The fund’s edge is supposed to come from seeing where the real bottlenecks sit before traditional markets price them in. What the filing shows is a tactical rotation inside that same long-term view. Aschenbrenner is staying aggressively long the broad AGI buildout while exiting or hedging the part of the supply chain that got the frothiest: semiconductors. At the same time he is adding heavily to power, energy, and flexible hosting plays — CleanSpark added over 10 million shares, CoreWeave another million-plus, with continued exposure to Riot, IREN, Applied Digital, and Bloom Energy. The signal is clear: semis are priced for perfection, while power and rapid-deploy compute capacity remain the scarce resources that will actually determine who captures margin as the buildout continues. Bitcoin miners that have quietly converted capacity to AI hosting now look especially well positioned under this lens. They already cleared the hardest hurdle — securing grid interconnection and power. That advantage is suddenly worth far more than a slightly better chip process node. This filing is sophisticated capital acknowledging that the game has moved downstream. The people who wrote the AGI timelines are now positioning for a world where power is the new oil and hosting flexibility is the real moat.

Ohh hi there @LebanonJon

The neoclouds quietly became the second tier of hyperscale AI infra. Same 100kW racks, same RoCE fabrics, same firmware nightmares — on a fraction of the team and balance sheet. That's the gap OCP's new Scaling AI Clusters at Neoclouds workstream exists to close, and FarmGPU is co-chairing it alongside Scaleway and Denvr Dataworks. A few signals from the first 19-operator community survey worth paying attention to: → Power sourcing and benchmarking are rated the most important 2026 problems (4.21 / 4.05 importance). → But the area where operators think OCP can move the needle most is open rack and pod reference patterns for 30–100kW (4.11 leverage). Hardware ≠ the bottleneck. Operational surface area is. → The top operator pain points: networking (RoCE/IB tuning, congestion control), management & automation (Day-2 ops, firmware), and interoperability (BMC/BIOS/telemetry consistency). Every neocloud is independently re-discovering the same gaps and filing the same tickets to the same OEMs. Concrete output already in motion: 1. A neocloud Redfish profile. Redfish works great on Dell, inconsistently everywhere else. Instead of 20 operators each begging OEMs separately, the workstream is drafting a collective, machine-validatable profile that codifies the BMC, firmware, TLS, telemetry, and event-subscription behaviors neoclouds actually need. One spec OEMs implement against. One checklist procurement validates. 2. FarmGPU is the first neocloud running an OCP-aligned backend fabric in production — the AI Training Fabric Reference Architecture we co-developed with Hedgehog and Celestica. Celestica DS5000 (51.2T) + SONiC + Hedgehog as the K8s-native control plane. 36 ready-to-deploy composition variants spanning OPG-64/128/256/512 → XOC-256/512/1024, air- and liquid-cooled, single- and dual-plane. Vendor-neutral, declarative, reproducible. Contributed back to OCP so nobody has to reinvent it. 3. OCP NVMe SSD spec → predictive failure engine. We standardize on the OCP Datacenter NVMe SSD spec because the telemetry surface is dramatically richer than vanilla NVMe. On top of that we built a fleet-wide predictive failure engine — scores every drive for NAND degradation, ECC escalation, recoverable/uncoverable error trends, surfaces replace/watch/monitor recommendations. Plan is to open-source it so other neoclouds can run it directly. The whole thesis: write down what neoclouds actually need → get OEMs to ship it through OCP → open-source the operator tooling that turns the spec into operational value. That compounds. Full writeup, survey results, hardware-management deep dive recording, and how to join the workstream: blog.farmgpu.com/ocp-scaling… 🌾🚜

Everyone benchmarks GPUs. Almost no one benchmarks the cluster. @lebanonjon on @SDxCentral: storage, fabric, and the boring stuff that decides whether your H100s actually earn their keep. sdxcentral.com/video-intervi…

🚜 The OCP just published the first real framework for DC-powered AI data centers. Most of the industry hasn't read it yet. On March 30, the OCP DCF Power Distribution sub-project — co-led by ABB and Eaton, with NVIDIA, Google, Microsoft, Amazon, AMD, Siemens, Schneider, Vertiv, and 180+ others on the workstream — shipped a white paper on Low Voltage DC power distribution for AI factories. This isn't a research curiosity. It's the architectural foundation for every NVIDIA 800V DC rack Jensen has been talking about, and the reason Mt. Diablo exists. The thesis in one sentence: AC has a 100-year head start, but it can't carry a 1+ MW rack without losing 5–10% to conversion stages and choking on 30→100% load swings in milliseconds. So the industry is rediscovering DC. The numbers worth stopping for: → ORv3 HPR power shelves went from 33 kW (v1, AC) → 72 kW (v2, AC) → 100 kW (v4, DC) in the same 1 OU footprint. A 39% density jump, almost entirely from deleting the AC/DC conversion stage. → Moving the BBU and CBU out of the IT rack into the Mt. Diablo sidecar reclaims 12–16 OU of compute volume. Another 27–36% of usable rack space, given back to GPUs. → AI training availability target: 99.9%. Inference: 99.999%. The catch nobody talks about: DC distribution cannot use UPS maintenance bypass. Redundancy has to live at the power block level. This rewrites how you think about N+1. → AI workloads ramp 30→100% in milliseconds. The CBU (capacitor backup) responds in sub-ms. AC architectures can't keep up. Grid operators have noticed. If your data center can't physically fit a Mt. Diablo / 800V DC rack, you can't run the next NVIDIA generation at full density. The 415V/480V room is already a stranded asset for hyperscale AI. Read the whitepaper: opencompute.org/documents/dc… 🌾🚜

Confidential Computing: Securing Data Where It's Most Vulnerable — In Use You encrypt data at rest. You encrypt data in transit. But when a GPU processes it — training a model, running inference — it sits in memory in plaintext. Anyone with hypervisor or BMC access can see it. Confidential computing closes that gap. The Problem: Data in Use Is Data Exposed Three data states exist: at rest (encrypted, solved), in transit (encrypted, solved), and in use (traditionally wide open). For AI workloads, the stakes are massive — model weights that cost $50M-$500M to train sit in GPU memory unprotected. Training data with PII, financial records, healthcare info — all decrypted during computation. Confidential computing uses hardware-based Trusted Execution Environments (TEEs) to encrypt and isolate data DURING computation. Not after. Not during transport. While it's being processed. CPU TEEs: The Foundation Intel TDX (Trust Domain Extensions) — Creates hardware-isolated VMs called Trust Domains. Entire VMs run with encrypted memory (AES-XTS-128 via MKTME with 28-bit MACs). The hypervisor can schedule VMs but cannot read their memory or modify their code. TDX Connect extends this to PCIe devices, enabling encrypted DMA between CPU and GPU TEEs. AMD SEV-SNP — Three generations deep. SEV (encryption only) → SEV-ES (+ register protection) → SEV-SNP (+ memory integrity). SEV-SNP prevents hypervisor replay and corruption of encrypted memory. SEV-TIO extends TEE protection to PCIe devices. ARM CCA — Realm Management Extensions create isolated Realms. Relevant for Ampere-based cloud and edge. CPU TEEs are necessary but not sufficient for AI. They protect host-side code and data, but GPUs process the most sensitive information — and GPU memory has historically been unprotected. GPU TEEs: The New Frontier NVIDIA's H100 was the first production GPU with confidential computing. The architecture: 1. CC-On Boot — GPU boots with hardware protections enabled. On-die silicon root of trust. Debug access, JTAG, and performance counters disabled. BMC restricted to anonymous health metrics only. 2. Secure Boot Chain — Signed firmware verified during boot. Firmware measurements stored in hardware registers. Chain of trust from silicon to software. 3. SPDM Key Exchange — In-CVM driver and GPU firmware establish an SPDM session via Diffie-Hellman. Both authenticate, exchange firmware measurements, and establish AES-256-GCM session keys for all CPU↔GPU data. 4. Data Transfer Modes: • Bounce Buffers — Shared encrypted pages, AES-256-GCM with rolling 96-bit IVs. Hypervisor sees ciphertext only. Works on Hopper and Blackwell. • TDISP/IDE (Blackwell + Granite Rapids) — GPU DMAs directly into CVM-private memory. No bounce buffers. IDE encrypts every PCIe TLP at line rate. Zero-copy, near-zero overhead, hardware end-to-end encryption. 5. GPU Memory Encryption — All HBM encrypted. Keys managed by GPU security processor, ephemeral, destroyed on teardown. 6. NVLink Encryption — Blackwell encrypts GPU-to-GPU traffic at 1.8 TB/s via AES-GCM. Critical for distributed training and large-model inference. 7. Remote Attestation — Cryptographically signed reports of firmware version, boot measurements, and configuration. Dual attestation: Intel TDX + NVIDIA GPU TEE providing independent verification. What It Protects — And What It Doesn't Protects against: Cloud operator insider access, compromised hypervisors (VM escape), physical access (PCIe analyzers, DIMM dumps, JTAG), co-tenant VMs, supply chain attacks (with attestation). Does NOT protect against: Side-channel attacks (raised bar, not eliminated), compromised app logic inside the TEE, output inference attacks (differential privacy is the countermeasure). The Trust Boundary • Trusted: Inside the CVM — your application, CUDA kernels, model weights, training data. Encrypted in CPU memory by TDX/SEV, encrypted in GPU memory by NVIDIA CC. Only you hold the keys. • Encrypted boundary: PCIe link + bounce buffers (or TDISP/IDE on Blackwell). All data AES-256-GCM encrypted. With TDISP/IDE, even shared memory intermediaries are eliminated. • Untrusted: Hypervisor, host OS, BMC, physical access. All see only ciphertext. Keys ephemeral, destroyed on teardown. Why Now 1. Regulation is enforceable. GDPR requires data protection during processing. HIPAA requires safeguards for PHI during computation. EU AI Act (enforceable August 2027) classifies models >10^25 FLOPS as systemic risk. 35+ countries have sovereign AI programs. This is procurement requirements, not theory. 2. Multi-party AI needs it. Training on combined datasets without exposing raw data requires TEEs. Alternatives (homomorphic encryption, SMPC) are 10-1000x slower. TEEs make multi-party AI commercially viable. 3. Model IP is a survival issue. Frontier training runs cost $50M-$500M. Without TEEs, any admin can dump model weights. CC converts IP protection from a trust relationship into a cryptographic guarantee. The Market • 75% of orgs adopting CC (IDC, Dec 2025), 18% in full production • Market: $12.28B (2025) → $54.92B (2030), 34.7% CAGR • Azure: only hyperscaler with GA confidential GPU VMs (SEV-SNP + H100, single GPU) • Google Cloud: TDX + H100, 3 zones, GA Aug 2025 • AWS: no GPU TEE capability (Nitro Enclaves can't access GPUs) • CoreWeave, Lambda, Crusoe, RunPod: zero CC capability Hardware's here. Stacks are maturing. Regulations are enforceable. The question isn't whether — it's how fast. 🌾

We partnered with @RunPod to build something the GPU cloud market hasn't had before: bare-metal B200 training performance with instant cluster spin-up, Docker-native, pay by the second. The "instant" part is real. You spin up a 6-node B200 HGX cluster the same way you'd launch a RunPod pod — minutes, not months. No procurement cycle. No contract. No minimum runtime. Same UX. Training-class hardware. Under the hood: • 800G open Ethernet fabric (Celestica DS5000 + @githedgehog) — 50% cheaper than Spectrum X, 400 GB/s GPU-to-GPU bandwidth • Solidigm PCIe 5.0 NVMe, 8 drives per node — 116 GB/s local storage, #1 in @SemiAnalysis_ ClusterMAX 2.0 • RoCEv2 + NVMe-oF, custom optics for CX7/BF3. OCP-compliant, fully open-spec. On NCCL: 390 GB/s AllReduce bus bandwidth on 32 B200s. 2.3x performance improvement at 16 MB message size — the exact range SemiAnalysis identifies as critical for real training workloads. Sub-50µs latency for small messages. Why does that matter? A network that's half as slow on AllReduce costs you 10% MFU on 70B parameter training and 15-20% on MoE models. That's real dollars, idle GPUs, wasted runs. We made NCCL performance the obsession, not an afterthought. The RunPod magic is the combination: full training performance inside a Docker container, on bare metal, with the same experience you'd use to launch a single GPU pod. No SLURM expertise. No ops team. Full cluster, browser to training job, in minutes. Two years building this with @RunPod. ClusterMAX #1 in storage. 390 GB/s NCCL. B200 bare metal, instantly available. blog.farmgpu.com/b200-cluste… 🌾

B200 bare metal performance. Spins up in minutes. Good news FarmGPU + @RunPod Instant Clusters: B200 HGX, 800G open Ethernet fabric, 8x PCIe 5.0. NVMe per node via @Solidigm No procurement queue. No hyperscaler tax. Just the hardware. Bad news we are sold out 😭 need more GPUs blog.farmgpu.com/b200-cluste…