Synthszr Charts — die großen AI-Marken im Wettkampf ums Podium
synthszr charts
nvidia

Nvidia · v200 · nvl 72 · 2× · vu le 02 juil. 2026

4
Momentum

The NVIDIA GB200 NVL72 is a liquid-cooled, rack-scale system that connects 36 Grace CPUs and 72 Blackwell GPUs via fifth-generation NVLink Domain into a single logical unit. The system is designed for Training and Inference of large language models in the trillion-parameter range and is delivered as an integrated rack by NVIDIA and partners such as Dell, HPE, and Supermicro. Cloud providers including CoreWeave, Oracle, Azure, and Google Cloud offer GB200 NVL72 instances commercially; individual purchases are typically made through system integrators or cloud contracts rather than at retail pricing.

Historique du momentum
18.08.16.11.

Fonctionnalités

Manufacturing Process (nm)TSMC custom 4NP process (4-nanometer-class, variant of N4P), 208 billion transistors per Blackwell GPU
LicenseProprietary NVIDIA hardware; software stack (CUDA, TensorRT-LLM, NeMo) under NVIDIA license terms, not an open-source product
PlatformNVIDIA Grace Blackwell platform: 36 Grace CPUs (Arm Neoverse V2) + 72 Blackwell GPUs connected via 5th-gen NVLink domain in a liquid-cooled rack
PriceNo fixed list price; integrator estimate for comparable GB200 NVL36 rack ~$1.8M, single GB200 Superchip ~$60,000–70,000 (estimates, not official NVIDIA pricing)
Compute Performance (FLOPS/TOPS)1,440 PFLOPS NVFP4 (sparse) / 720 PFLOPS dense; 720 PFLOPS FP8/FP6; 360 PFLOPS FP16/BF16; 130 TB/s NVLink bandwidth per rack
Release DateFirst system shipments to customers (incl. Microsoft, Oracle, Meta) in December 2024; CoreWeave was first cloud provider with general availability on February 4, 2025
Memory13.4 TB HBM3e GPU memory at 576 TB/s bandwidth per rack; plus 17 TB LPDDR5X CPU memory at 14 TB/s
AvailabilityWidely available since early 2025 via cloud providers (CoreWeave, Oracle Cloud, Azure, Google Cloud) and system integrators such as Dell, HPE, Supermicro

Plus de produits dans cette catégorie: Matériel d'inférence IA

Preuves (2)

Company Analysis: Nvidia

As of 07/11/2026
HOLDSynthszr Vote

Data-driven balance: NVIDIA’s latest reported fundamentals are exceptionally strong (Q2 FY2027 revenue $96.2B, +18% QoQ and +106% YoY; gross margin ~75%), and filings indicate Blackwell is already the majority of revenue—both supportive of continued platform leadership. (nvidianews.nvidia.com) However, management commentary highlights supply and memory constraints as key swing factors for near-term margins and shipment timing, which can drive earnings revisions and valuation volatility over the next 1–2 quarters. (investor.nvidia.com) With forward valuation appearing lower than prior years but still highly sensitive to growth/margin revisions, HOLD is the most disciplined stance until the next earnings update confirms that guidance converts into realized shipments and stable margins. (uk.marketscreener.com)

Summary

NVIDIA is an AI infrastructure company whose core business model is to sell accelerated computing platforms (GPU/CPU/DPUs plus networking, systems, and software) that enable training and inference for large-scale AI, along with adjacent graphics and edge workloads. Its core competencies are (1) leading GPU architecture and packaging, (2) tightly integrated full-stack platform delivery (hardware + CUDA software ecosystem + libraries), and (3) high-velocity product cadence that keeps performance-per-watt and total cost of ownership advantages in front of customers. Market position remains dominant in AI data center acceleration, with competitive advantages rooted in software lock-in (CUDA and a broad developer ecosystem), platform breadth (GPU + networking + systems), and deep integration with hyperscalers and OEMs. Recent disclosures indicate Blackwell is the “vast majority” of revenue, underscoring NVIDIA’s ability to transition architectures at scale while maintaining premium pricing and high utilization of supply. (investor.nvidia.com) In the most recent reported quarter (Q2 FY2027, quarter ended July 26, 2026), NVIDIA reported revenue of $96.2B, up 18% sequentially and up 106% year-over-year, with gross margin around 75% (company-reported). (nvidianews.nvidia.com) Management commentary emphasized strong hyperscaler demand and continued supply constraints (notably memory), which can influence near-term margin dynamics even as revenue scales. (investor.nvidia.com) Separately, filings highlight that China exposure for certain data center products has been minimal (e.g., Hopper shipments to China in Q2 FY2027 were <1% of Data Center revenue), consistent with export-control-related demand reallocation. (investor.nvidia.com) Valuation and key metrics: consensus valuation snapshots show a materially lower forward P/E versus prior years as earnings have expanded; one market consensus page indicates P/E stepping down into the high-teens for FY2027 and low-teens for FY2028 (calendarized by NVIDIA’s January fiscal year). (uk.marketscreener.com) A separate market commentary source frames NVIDIA around mid-20s forward P/E versus large-cap AI-exposed peers, highlighting that relative valuation depends heavily on the durability of AI capex and supply normalization. (01.co) Outlook (short- to medium-term): near-term performance is primarily a function of (1) Blackwell supply ramp and memory availability, (2) hyperscaler and “sovereign AI” deployment schedules, and (3) margin sensitivity to component costs and mix. Management’s Q2 FY2027 outlook for Q3 FY2027 (revenue and margin guidance) and the earnings-call narrative point to continued strong demand with supply as the binding constraint, implying execution and supply-chain throughput are the key swing factors over the next 6–18 months. (nvidianews.nvidia.com) For EUR framing, the ECB reference rate on Nov 6, 2026 was 1 EUR = 1.1165 USD (useful for translating USD-denominated valuation/financials). (ecb.europa.eu)

Key Takeaways

  1. NVIDIA’s current growth profile is still driven overwhelmingly by AI data center platforms; Q2 FY2027 revenue was $96.2B (+18% QoQ, +106% YoY), indicating demand remains well above historical semiconductor cycles. (nvidianews.nvidia.com)
  2. Gross margins remain structurally high (around 75% reported in Q2 FY2027), but management commentary and filings indicate component constraints (notably memory) can pressure near-term margin even as revenue grows. (nvidianews.nvidia.com)
  3. Blackwell is already the majority of revenue per filings, reducing transition risk versus prior architecture shifts and supporting platform continuity for hyperscalers and OEMs. (investor.nvidia.com)
  4. China-related data center compute exposure appears limited in recent reporting (e.g., Hopper shipments to China <1% of Data Center revenue in Q2 FY2027), which reduces single-region dependency but also caps upside if restrictions persist. (investor.nvidia.com)
  5. Consensus valuation has compressed versus earlier years as earnings scaled; market consensus pages show forward P/E in the high-teens for FY2027 and low-teens for FY2028, making the stock’s risk/reward increasingly tied to sustaining AI capex and NVIDIA’s supply execution. (uk.marketscreener.com)

Action Ideas

BUY

Time the position around fundamental catalysts rather than price momentum: the latest reported quarter (Q2 FY2027) showed very strong sequential and year-over-year revenue growth and high gross margins, while management messaging frames supply (including memory) as the key limiter. A BUY is supported if you underwrite (a) continued hyperscaler deployment schedules and (b) NVIDIA’s ability to convert backlog into shipments without a material margin reset. Use EUR translation via ECB reference rates for portfolio-level risk budgeting (USD exposure).

Horizon: 12 mo.

HOLD

Maintain exposure if you already own shares and require confirmation that guidance translates into realized shipments. The Q2 FY2027 results and Q3 outlook indicate strong demand, but the binding constraint is supply and component costs; a HOLD stance is appropriate if you want to see evidence of sustained margin stability through the next 1–2 quarters while Blackwell volumes expand.

Horizon: 6 mo.

SELL

Reduce exposure if your mandate cannot tolerate estimate-risk and multiple compression risk. Even with strong fundamentals, the equity’s performance can be dominated by changes in consensus growth assumptions and near-term margin guidance. A SELL is most defensible for investors who require lower earnings-variance and less dependence on a single secular theme (AI infrastructure buildout).

Horizon: 3 mo.

Contrarian Insights

  • • Consensus often frames NVIDIA’s near-term risk as demand-driven (AI capex ‘overbuild’). A contrarian read from management’s recent disclosures is that supply (including memory) remains the binding constraint; if true, near-term downside may be more about margin variability than revenue collapse. (investor.nvidia.com)
  • • Many investors treat China as a large latent upside lever. Recent filings show China-related Hopper shipments were <1% of Data Center revenue in Q2 FY2027, implying the current run-rate is already largely ex-China; the contrarian implication is that incremental China reopening may be less central to the base case than commonly assumed, while policy volatility remains a risk factor. (investor.nvidia.com)

Sources (7)

Subscribe free. Unsubscribe the second it sucks.

High-signal news across AI, business, UX, and tech. Every morning.