Feb 26, 2026 · 24-Hour AI Briefing: NVIDIA frames Agentic AI as the next inflection, AMD teams up with Nutanix for enterprise delivery, IonQ’s “profitable quarter” raises questions
The last 24 hours offered a clear snapshot of where AI is headed next: NVIDIA continues to monetize an end-to-end infrastructure stack with software-like margins while pushing the narrative toward Agentic AI; AMD is trying to break out by winning the enterprise deployment layer with Nutanix; and IonQ delivered a headline financial surprise that still needs careful separation between operating progress and non-recurring accounting effects.
1. NVIDIA reports FY2026 Q4 revenue of $68.1B (+73% YoY); datacenter revenue $62.3B, gaming $3.7B; GAAP/non-GAAP gross margin 75.0%/75.2%
Jensen Huang says the inflection point for Agentic AI has arrived.
Commentary:
Datacenter contributed $62.3B, about 91.4% of total revenue. From GPUs (Blackwell), high-speed interconnect (NVLink), networking (Spectrum-X), software (CUDA, AI Enterprise), to full systems (DGX, GB200 Superpod), NVIDIA is one of the few vendors offering a tightly optimized, end-to-end AI infrastructure stack—one reason it can sustain very high gross margins.
Agentic AI isn’t about better chat. It’s about models entering real operational loops: decomposing tasks, calling tools, executing actions, and learning from feedback. That shifts demand from “train once, infer many” toward more complex inference patterns—longer chains, higher call frequency, tighter latency constraints—driving more inference compute and heavier systems investment. For NVIDIA, that’s a strong tailwind beyond the training-only era.
The risk is concentration and geopolitics: if more than half of datacenter revenue is tied to a small set of hyperscalers (Amazon, Microsoft, Google, Meta), any one of them slowing capex can bend the near-term growth curve. Meanwhile, datacenter exposure tied to China remains uncertain under ongoing geopolitical constraints.
2. AMD and Nutanix to co-develop an open full-stack AI infrastructure platform; AMD invests $150M at $36.26/share and provides up to $100M for joint engineering and go-to-market
Commentary:
With CUDA’s ecosystem advantage still enormous, AMD needs “breakout vectors” beyond raw hardware. The core bet here is enterprise delivery: using Nutanix’s HCI platform as a control plane for simpler AI cluster deployment, scheduling, and operations, while hardening ROCm integration across common frameworks (PyTorch, TensorFlow) and Kubernetes-based environments.
Nutanix claims a base of 20k+ customers, largely non-internet enterprises. Its reputation for “out-of-the-box” operations helps address AMD’s gaps in enterprise software ecosystem and channel reach. In practice, enterprises don’t buy GPUs—they buy something that boots, runs, upgrades, monitors, and stays compliant.
The real question is whether “open” translates to acceptable enterprise experience. Enterprises want to avoid lock-in, but they also fear fragmentation. Without strict certification matrices, version governance, and observability, “open” can become another word for integration pain. If the partnership can deliver “open, but not chaotic,” it could become a meaningful route for AMD into enterprise AI.
3. IonQ reports Q4 revenue of $61.9M and EPS of $1.93 vs. analysts expecting a -$0.47 loss
Commentary:
The shift from “burning cash” to “a profitable quarter” is attention-grabbing: $61.9M in Q4 revenue approaches the scale of the first three quarters combined (roughly $66M), which can happen in project-based businesses—but the EPS swing from an expected loss to a large profit demands scrutiny. EPS can look great without implying the core business is self-funding, especially when one-time items or fair-value accounting can dominate reported earnings.
Quantum computing is still early in commercialization. Investors should focus on delivery and technical roadmaps: performance progression, error-rate reduction, scalability, and whether technical advantages translate into repeatable enterprise contracts and growing cloud usage. Until large-scale error correction and practical algorithms mature, “sudden profitability” stories deserve extra caution.