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

Cosmos 3 Super

Nvidia · v3 · super · 8× · last seen Jul 24, 2026

Momentum

Cosmos 3 Super is the most powerful variant of NVIDIA's Cosmos-3 platform, an open, omnimodal Foundation Model for Physical AI. It is a 64-billion-parameter model (32B Reasoning tower plus 32B Generator tower) in a Mixture-of-Transformers architecture that processes text, image, video, audio, and robotic actions in a unified system. It was released on May 31, 2026 and officially presented on June 1, 2026 at GTC/Computex Taipei, targeting datacenter GPUs (Hopper/Blackwell) and primarily serving Post-Training of robotics and autonomous vehicle models with high physical accuracy. Model weights are freely available under the open OpenMDW-1.1 license on Hugging Face; a fixed commercial list price

Momentum trend

11.06.09.09.

Features

Fine-Tuning	Open post-training scripts for domain adaptation available; explicitly designed for post-training robotics and AV models
Generation Time	Approx. 55 seconds for a 50-step video generation on H200 GPU configuration (2x H200: approx. 3 minutes)
License	OpenMDW-1.1 (Linux Foundation Open Model Development Weights License), commercial use permitted
Max Resolution	720p (also supports 480p and 256p)
Max Video Length	5 to 300 frames, default 189 frames (about 7.9s at 24 FPS); limited to 1–5 seconds via DeepInfra API
Platform	Runs on NVIDIA Hopper and Blackwell datacenter GPUs (H100, H200, B200); available via Hugging Face, build.nvidia.com, NVIDIA NIM, GitHub, DGX Cloud
Price	Model weights free (open-weight) via Hugging Face; third parties like DeepInfra charge usage-based, resolution-dependent API fees (no fixed NVIDIA list price found)
Release Date	May 31, 2026 (Hugging Face/GitHub release), officially announced June 1, 2026 at GTC Taipei

More products in this category: 3D & World Generation

Sources (8)

Company Analysis: Nvidia

As of 29/08/2026

HOLDSynthszr Vote

Data-driven balance of factors supports HOLD at current levels (≈$217.55, ≈€187.79): fundamentals and guidance remain very strong (Q2 revenue $96.221B; Q3 guide ~$108B ±2%; gross margin guided ~74% ±50 bps), but the stock’s valuation (trailing P/E ~27x) implies high expectations and increases sensitivity to any execution or guidance variance. Maintaining exposure is justified while monitoring (1) supply ramp execution and commitments, (2) margin trajectory versus guidance, and (3) regulatory/export constraints already reflected in the China assumption.

Summary

NVIDIA is a fabless semiconductor and software platform company whose business model is centered on designing accelerated computing hardware (GPUs, networking, and systems) and monetizing a full-stack software ecosystem (CUDA, libraries, AI frameworks, and enterprise software) that increases customer switching costs and expands attach rates. The company’s core competency is end-to-end AI infrastructure enablement: silicon, interconnect, systems, and developer tooling that together optimize performance-per-watt and time-to-train/time-to-inference for large-scale AI workloads. Market position remains dominant in AI accelerators and tightly coupled networking/system design for hyperscale and enterprise AI deployments. Competitive advantages are reinforced by (1) a large installed base and developer ecosystem, (2) rapid product cadence (e.g., Blackwell platform referenced in recent commentary), and (3) the ability to sell integrated platforms rather than discrete chips, which supports pricing power and high gross margins. In the most recent reported quarter (fiscal 2027 Q2, ended July 26, 2026), NVIDIA posted record GAAP revenue of $96.221B (+106% YoY) and Data Center revenue of $89.0B (+117% YoY). Gross margin was 75.0%, and external coverage highlighted operating margin expansion to ~66% (company and third-party recap). Management guided fiscal 2027 Q3 revenue to approximately $108B (±2%) and guided gross margin to ~74.0% (±50 bps). Management also reiterated that its forward outlook assumes no Data Center compute revenue from China due to licensing uncertainty, which is a material geographic constraint embedded in guidance. On the earnings call, management indicated a preliminary expectation for fiscal 2028 revenue growth of ~70% YoY, framing the outlook as supply-constrained. Valuation and key metrics (USD-based) show NVDA trading at ~27.3x trailing P/E at $217.55 per share (Aug 29, 2026). Converting the latest price to EUR using a recent EUR/USD reference (~1.1585 USD per EUR) implies ~€187.79 per share. Near-term outlook is anchored by (a) continued hyperscaler capex intensity for AI infrastructure, (b) platform transitions and supply ramp execution, and (c) sustained gross margin in the mid-70% range per guidance. Medium-term performance sensitivity remains high to customer concentration, supply-chain execution at unprecedented scale, and regulatory/export constraints already reflected in management’s China assumption.

Key Takeaways

Fiscal 2027 Q2 results were materially above prior-year levels: $96.221B revenue (+106% YoY) with Data Center at $89.0B (+117% YoY).
Profitability remains exceptional for a semiconductor company: 75.0% gross margin in Q2 and operating margin around the mid-60% range per third-party recap.
Management guided fiscal 2027 Q3 revenue to ~$108B (±2%) and gross margin to ~74.0% (±50 bps), indicating continued high growth with modest margin normalization versus Q2.
Guidance explicitly assumes no Data Center compute revenue from China due to export/import licensing uncertainty, embedding a meaningful regulatory constraint into forward expectations.
At ~$217.55 (≈€187.79), trailing P/E is ~27.3x, implying the market is pricing sustained high earnings power and continued execution against aggressive growth expectations.

Action Ideas

BUY

For investors with a 12–24 month horizon, NVIDIA’s latest reported fundamentals and guidance support a constructive stance: Q2 revenue of $96.221B (+106% YoY) and Q3 revenue guidance of ~$108B (±2%) indicate continued demand strength and supply ramp progress. Gross margin guidance (~74%) remains structurally high, consistent with platform-level pricing power and software/ecosystem leverage. The explicit exclusion of China Data Center compute revenue in guidance reduces the risk of near-term estimate resets from that region, shifting the focus to execution and supply scaling in other geographies.

Horizon: 18 mo.

HOLD

For investors already positioned, HOLD is supported by strong near-term visibility (Q3 revenue guidance ~$108B) and sustained high gross margins, balanced against a valuation that already reflects substantial earnings power (trailing P/E ~27x). This stance fits portfolios that want continued exposure to AI infrastructure growth while monitoring whether growth and margins remain consistent as the revenue base scales further and as supply commitments expand.

Horizon: 9 mo.

SELL

For risk-managed or valuation-disciplined investors with a 3–6 month horizon, SELL/trim can be justified after a period of extraordinary growth and profitability, given the stock’s sensitivity to quarterly guideposts and the possibility of volatility around supply ramp execution and customer ordering patterns. With trailing P/E ~27x and expectations for continued outsized growth, the near-term risk/reward may be less favorable for investors who cannot tolerate drawdowns tied to guidance cadence and regulatory headlines.

Horizon: 6 mo.

Contrarian Insights

• China exposure risk may be less of a near-term earnings shock than commonly assumed because management’s forward outlook already assumes no Data Center compute revenue from China; incremental negative impact would require restrictions to broaden beyond what is already embedded in guidance.
• The primary near-term risk may be operational scaling (supply commitments and delivery execution) rather than demand: management framed the outlook as supply-constrained and third-party recap highlighted rapidly rising supply commitments, implying execution discipline could be a more important driver of variance than end-market appetite.