Telemetries

6 of 6 insights

–
Event

NVIDIA invests $2 billion in Synopsys to accelerate GPU EDA

  • Main point
    • NVIDIA is making a $2 billion equity investment in Synopsys and partnering to port physics-based EDA simulations from CPUs to NVIDIA GPUs via CUDA, aiming to remove chip verification bottlenecks.
    • Synopsys, one of the two dominant EDA vendors alongside Cadence, currently runs most verification workloads on CPUs, where simulations can take weeks or months.
    • Porting these workloads to GPUs requires a fundamental algorithmic and codebase overhaul to exploit massive thread-level parallelism and CUDA, rewriting code accumulated over decades.
    • Synopsys’ CEO reports that early CUDA-based products already deliver 10–20x speedups, reducing multi-week jobs to hours, and notes that verification compute is the primary bottleneck in complex chip design.
    • Large language models are expected to help automate code migration and optimization, creating a feedback loop where GPUs help generate software that further increases GPU demand.
    • The partnership is nonexclusive, but strategically important for NVIDIA as faster GPU-based verification supports its move to an annual major-chip release cadence and lowers industry-wide design and verification costs.
NVIDIASynopsysCadence
AI-driven Semiconductor Manufacturing

Dec 14, 2025, 6:59 AM

Confirming

GPU-accelerated EDA positions Synopsys for broader digital twin market

  • Main point
    • The NVIDIA–Synopsys partnership aims to extend EDA from serving only the semiconductor industry to enabling digital twin-based design across virtually all physical products, vastly expanding Synopsys’ addressable market.
    • Jensen Huang argues that every future product will first be designed as a digital twin, turning EDA and CAD tools into a gateway to the R&D budgets of most of the world’s industrial GDP.
    • He notes that scientific simulation supercomputing has already shifted from 90% CPU-only in 2016 to 90% NVIDIA-accelerated systems today, and expects a similar transition in EDA toward CUDA-accelerated workflows.
    • Realizing this vision requires multi-year reformulation of Synopsys’ algorithms to run efficiently on CUDA, similar to the seven-year journey for NVIDIA’s cuLitho.
    • The collaboration is nonexclusive, allowing Synopsys to keep working with other chip partners, and NVIDIA to continue partnering with other EDA and CAD vendors like Cadence, Siemens, and Dassault, reinforcing a broader GPU-centric ecosystem for digital twins.
AI-driven Semiconductor ManufacturingDigital Twin-Based Product Design

Dec 14, 2025, 6:59 AM

Event

Marvell acquires Celestial AI to build photonic fabric interconnect

  • Main point
    • Marvell is acquiring Celestial AI to add a photonic fabric platform for next-generation AI data center scale-up interconnects, deepening its strategic pivot toward data center and AI markets.
    • The deal follows a series of portfolio moves since 2019 (divesting WiFi, acquiring Avera, Aquantia, Inphi, Innovium) that helped drive 41% year-over-year revenue growth this quarter and raised data center to 74% of company revenue.
    • Celestial AI provides a photonic fabric platform and a first-generation photonic fabric chiplet delivering 16 Tbps bandwidth—10x current 1.6T ports—offering higher bandwidth, longer reach, and more than 2x power efficiency compared to copper interconnects.
    • Its technology supports nanosecond-class latency and thermal stability suitable for co-packaging with high-power XPUs and switches, enabling direct in-package optical connections and freeing die-edge area for more HBM.
    • Celestial AI is already deeply engaged with multiple hyperscalers and has won a major design with one of the world’s largest hyperscalers; Q&A indicates Amazon AWS will be the first customer, deploying PF chiplets in future Trainium-based multi-rack AI training clusters.
    • Marvell expects Celestial AI revenue to reach a $500 million annualized run rate by Q4 fiscal 2028 and $1 billion by Q4 fiscal 2029, with the acquisition expected to close in the first quarter of next year subject to U.S. regulatory review.
MarvellCelestial AIInfineonAveraAquantiaInphiInnoviumAmazonCredo
AI-driven Semiconductor ManufacturingOptical Scale-Up Interconnects for AI Data Centers

Dec 14, 2025, 6:59 AM

Confirming

Marvell faces Broadcom competition but AI optical TAM expands

  • Main point
    • HSBC expects Marvell’s AI optical business to keep growing strongly despite intensifying competition from Broadcom, driven by expanding AI infrastructure demand and new 1.6T DSP adoption.
    • Marvell currently maintains market leadership in PAM4 DSPs, with strong demand for its 800G solutions and anticipated acceleration in 1.6T deployments over the coming quarters.
    • Broadcom has caught up with its own 800G and 1.6T DSP offerings, which is likely to slow Marvell’s growth rate even as absolute revenues continue rising.
    • HSBC forecasts Marvell’s AI optical business year-on-year growth decelerating from 54% in FY26 to 38% in FY27 and 20% in FY28, but still above Street expectations of 20% in FY27 and a 3% contraction in FY28.
    • The bank’s bullish view is based on increasing AI infrastructure demand beyond traditional hyperscalers into enterprises and sovereign AI projects, which should expand the total addressable market for high-speed optical interconnect semiconductors.
BroadcomHSBC
AI-driven Semiconductor Manufacturing

Dec 14, 2025, 6:59 AM

Event

Siemens and GlobalFoundries partner on AI-driven semiconductor manufacturing

  • Siemens and GlobalFoundries announced a collaboration to deploy AI-driven manufacturing aimed at strengthening global semiconductor supply.
    • The partnership is framed as using artificial intelligence, including generative AI, to enhance manufacturing efficiency and resilience in the semiconductor supply chain.
    • Siemens brings its position as a leader in industrial AI and its experience applying AI to real-world industrial applications to this collaboration.
    • The initiative reflects a broader move to combine real and digital technologies in factories to improve performance, reliability, and sustainability.
SiemensGlobalFoundries
AI-driven Semiconductor Manufacturing

Dec 14, 2025, 6:56 AM

Event

Siemens reports fiscal 2025 revenue and net income performance

  • Siemens reported substantial fiscal 2025 financial and employment figures, underscoring its global industrial scale.
    • For fiscal 2025, which ended on September 30, 2025, the Siemens Group generated revenue of €78.9 billion and net income of €10.4 billion.
    • As of September 30, 2025, Siemens employed around 318,000 people worldwide on the basis of continuing operations.
    • These figures highlight Siemens’ position as a large, diversified technology company active across industry, infrastructure, mobility, and healthcare markets.

Dec 14, 2025, 6:56 AM