返回新聞
產業AI Understanding 簡報

Navitas 將以高達 2.328 億美元收購 Claros,以促進人工智慧電力輸送

Navitas Semiconductor 宣布達成收購 Claros Inc. 的最終協議,增加垂直供電和整合穩壓器技術,旨在克服下一代人工智慧加速器的電源牆。

4 min readRead the linked source
Source-provided image accompanying Navitas to acquire Claros for up to $232.8 million to boost AI power delivery
來源參考來源記錄
出版商
datacenters.economictimes.indiatimes.com
來源連結
datacenters.economictimes.indiatimes.comhttps://datacenters.economictimes.indiatimes.com/news/ai-compute-infrastructure/navitas-to-acquire-claros-for-ai-data-centre-power-delivery/134554607
來源類型
連結來源-主要來源狀態尚未確定。
背景60 秒內了解這一點

從這裡開始

關鍵術語

GAN(生成對抗網路)
生成器和鑑別器相互訓練的生成設定。
計算
訓練和運行模型所需的處理資源,通常以 FLOPS 或 GPU 小時來衡量。
測試一下自己人工智慧測驗的未來

發生了什麼事

Navitas Semiconductor Corp. signed a definitive agreement to acquire Claros Inc. for up to approximately $232.8 million, based on Navitas’ closing stock price on Aug. 21, 2026. The deal will bring Claros’ vertical power delivery (VPD) and integrated voltage regulator (IVR) technologies into Navitas’ gallium‑nitride (GaN) and silicon‑carbide (SiC) product portfolio. Navitas says the combined offering will move power conversion from board‑level modules to within millimeters of the AI xPU (GPU, CPU, TPU, NPU), reducing impedance, improving efficiency, and enabling higher density in megawatt‑scale server racks.

Navitas Semiconductor, a leader in GaNFast gallium‑nitride and GeneSiC silicon‑carbide power semiconductors, entered into a definitive agreement to acquire Claros Inc., a specialist in vertical power delivery and integrated voltage regulation. The transaction value is tied to Navitas’ closing stock price on Aug. 21, 2026, with a maximum of about $232.8 million.

Claros’ technology stacks power conversion, drive, control, and passive components into a compact package that can be placed directly beneath or inside the AI accelerator chip package. This architecture shortens power‑travel distance to a few millimeters, delivering ultra‑fast transient response, lower impedance, and higher efficiency at sub‑volt levels—attributes essential for next‑generation AI workloads.

Navitas’ President and CEO Chris Allexandre said the acquisition will “break the AI infrastructure power wall, advancing the entire power chain from grid‑to‑xPU.” Claros co‑founder Dan Kultran highlighted that the close‑to‑chip approach can enable higher density, lower heat generation, and reduced operating costs for AI data‑center deployments.

來源詳情: datacenters.economictimes.indiatimes.com ↗

為什麼這很重要

Power delivery is a growing bottleneck for AI . Modern xPUs require thousands of amps and sub‑nanosecond response times, which traditional voltage regulator modules cannot reliably provide. By stacking conversion, drive, and control directly under or inside the chip package, Claros’ VPD and IVR solutions promise to cut power‑travel distance to a few millimeters, lowering losses, heat, and overall operating cost. Navitas projects that the acquisition will more than double its addressable market by 2030 to over $8 billion, adding roughly $3.5 billion from the fast‑growing VPD/IVR segment. The move deepens Navitas’ ties with hyperscalers and AI‑power platform providers, positioning the company as a key supplier for the next wave of AI infrastructure that must scale beyond current power‑wall limits.

The AI market is rapidly expanding, with hyperscalers building ever larger server farms that consume megawatts of power. Traditional VRM designs, which distribute power laterally across boards, struggle to meet the amperage and response‑time demands of modern AI accelerators, creating a “power wall” that limits performance scaling.

By integrating VPD and IVR directly at the chip level, Navitas aims to alleviate this bottleneck, allowing AI processors to operate at higher frequencies and densities without proportionally increasing power loss or cooling requirements. This could translate into more efficient data‑center operations and lower total‑cost‑of‑ownership for AI workloads.

Financially, Navitas expects the acquisition to more than double its addressable market by 2030, adding an estimated $3.5 billion from the VPD/IVR segment. The deal also strengthens Navitas’ position with hyperscalers, who are actively seeking power‑efficient solutions to sustain AI growth.

Interactive Mechanism

互動機制:它實際上是如何運作的

以互動方式探索這項發展背後的基礎技術。

Model Parameter Size:8B Parameters
VRAM Required5.5 GBGPU memory footprint
Target HardwareMacBook / Single GPUDeployment tier
Privacy100% Air-GappedLocal device capability
Core takeaway: Small, quantized models (3B–8B) now run directly inside smartphones and laptops with complete data privacy, while mammoth 400B+ models remain the domain of datacenter clusters.
互動式概念檢查+10 Points
Future of AI Quiz

What should a useful AI forecast state?

接下來看什麼

Key indicators to monitor include: (1) integration progress of Claros’ VPD/IVR modules into Navitas’ GaN and SiC product lines; (2) adoption by major hyperscalers or AI accelerator manufacturers, especially in upcoming 2027‑2028 server generations; (3) any announced pricing or availability timelines for the combined technology; and (4) regulatory scrutiny or supply‑chain impacts given the strategic importance of AI‑ power hardware.

Integration timeline: Navitas has indicated that the combined technology will start delivering value from 2028/2029 onward. Observers should watch for product roadmaps and sample shipments in the next 12‑18 months.

Customer uptake: Early adoption by major AI accelerator manufacturers (e.g., NVIDIA, AMD, Google) or hyperscale cloud providers will be a key signal of market acceptance.

Pricing and availability: The announcement did not disclose pricing or immediate availability. Any future announcements on cost structures or volume‑based pricing will affect the competitive landscape.

Supply‑chain and regulatory factors: As power semiconductors become strategic components for AI infrastructure, any export controls or geopolitical tensions could impact rollout.

相關指引和測驗

AI 的未來人工智慧代理人工智慧模型解釋測試你所知道的—嘗試免費的人工智慧測驗在我們的詞彙表中尋找人工智慧術語關注 AI 資金追蹤器
覺得有用嗎?