NVIDIA Vera Rubin/VR200 NVL72 的主要更新
（結(jié)合我最新的供應(yīng)鏈調(diào)查和英偉達(dá)CEO黃仁勛在2026年國(guó)際消費(fèi)電子展上的主題演講）
1. NVIDIA 已將 AI 服務(wù)器 VR200 NVL144（基于芯片）更名為 VR200 NVL72（基于封裝）。我的供應(yīng)鏈調(diào)查顯示，VR200 NVL72 將提供兩種功耗配置：Max Q 和 Max P。
? Jensen 引用了 NVL72，而不是之前使用的 NVL144 命名。
? Max Q 和 Max P 采用相同的硬件設(shè)計(jì)。
? Max Q GPU/機(jī)架功耗（TGP/TDP）：~1.8/190（kW）。
?最大 P GPU/機(jī)架功率（TGP/TDP）：~2.3/230（kW）。
?兩者均明顯高于 GB300 NVL72 的 1.4/140 (kW)。
?提供兩種功率模式可提高數(shù)據(jù)中心電力限制下的部署靈活性。這也表明 NVIDIA 越來(lái)越重視實(shí)際物理基礎(chǔ)設(shè)施的限制，并將這些限制納入產(chǎn)品規(guī)格中。
2. VR200 NVL72 升級(jí)了 MaxQ 和 MaxP 的 GPU 散熱解決方案。
?兩者都使用微通道冷板（MCCP）和鍍金蓋。
?市場(chǎng)一直期待GPU TGP達(dá)到2.3kW時(shí)會(huì)轉(zhuǎn)向微通道蓋（MCL）；然而，MCL的大規(guī)模生產(chǎn)不太可能在2027年下半年之前實(shí)現(xiàn)。
3. 得益于 GPU、HBM 和 NVLink 交換機(jī)的升級(jí)，VR200 NVL72 的訓(xùn)練/推理 AI 計(jì)算能力約為 GB300 NVL72 的 3.5 倍/5 倍，從而導(dǎo)致機(jī)架級(jí)電源需求急劇增加。
? VR200 NVL72 將電源架升級(jí)為 3*3U 110 kW（6*18.3 kW 電源），而最常見的 GB300 NVL72 配置為 8*1U 33 kW 電源架（6*5.5 kW 電源）。
? VR200 NVL72 電源架采用 3+1 冗余設(shè)計(jì)。
? VR200 NVL72 將功率鞭狀電流額定值提高到 100A（而 GB300 NVL72 為 60A），提高了數(shù)據(jù)中心電源基礎(chǔ)設(shè)施的要求（例如，母線槽和分接盒）。
4. VR200 NVL72 更加依賴液冷散熱。
?計(jì)算托架和 NVSwitch 托架均采用無(wú)風(fēng)扇設(shè)計(jì)。
?機(jī)架技術(shù)冷卻系統(tǒng)流量（TCS 流量）幾乎翻了一番(~+100%) ，與 GB300 NVL72 相比，有利于 CDU、歧管、冷板和快速斷開裝置 (QD) 的規(guī)格和/或數(shù)量升級(jí)。
?機(jī)架氣流需求比 GB300 NVL72 下降約 80%（以 CFM 為單位）。
5. VR200 NVL72 計(jì)算托架首次采用中板，實(shí)現(xiàn)了真正的無(wú)電纜設(shè)計(jì)。
?關(guān)鍵中平面規(guī)格包括 44 層(22+22) 、M9 CCL（EM896K3）和大約 420×60 毫米的尺寸。
? Jensen 指出，這種設(shè)計(jì)可以將計(jì)算托架組裝時(shí)間從大約 2 小時(shí)縮短到大約 5 分鐘（與 GB300 NVL72 相比）。
6. Rubin CoWoS 預(yù)計(jì)到 2026 年將生產(chǎn) 30 萬(wàn)至 35 萬(wàn)片晶圓，預(yù)計(jì)將于 2026 年第一季度初開始試生產(chǎn)，并于 2026 年第二季度末開始量產(chǎn)。VR200 NVL72 機(jī)架組件預(yù)計(jì)將于 2026 年第三季度末進(jìn)入量產(chǎn)階段；考慮到良率提升，預(yù)計(jì) 2026 年下半年機(jī)架出貨量約為 5000 至 7000 臺(tái)。
7. 當(dāng)機(jī)架功率接近約 200 kW 時(shí)，54V 配電在空間（母線/電纜）和轉(zhuǎn)換效率方面開始面臨材料限制。因此，VR200 NVL72 可以被視為 Oberon 機(jī)架架構(gòu)的最后一代產(chǎn)品。為了應(yīng)對(duì) AI 計(jì)算持續(xù)擴(kuò)展帶來(lái)的機(jī)架功率需求增長(zhǎng)，NVIDIA 預(yù)計(jì)將過(guò)渡到支持 800V 高壓直流 (HVDC) 的下一代 Kyber 機(jī)架設(shè)計(jì)。
原文參考
Key Updates on NVIDIA Vera Rubin/VR200 NVL72
(Integrating my latest supply chain checks and NVIDIA CEO Jensen Huang’s CES 2026 keynote)
1. NVIDIA has renamed the AI server VR200 NVL144 (die-based) to VR200 NVL72 (package-based). My supply chain checks indicate VR200 NVL72 will be offered in two power profiles: Max Q and Max P.
? Jensen referenced NVL72, rather than the previously used NVL144 naming.
? Max Q and Max P share the same hardware design.
? Max Q GPU/rack power (TGP/TDP): ~1.8/190 (kW).
? Max P GPU/rack power (TGP/TDP): ~2.3/230 (kW).
? Both are meaningfully higher than GB300 NVL72 at 1.4/140 (kW).
? Offering two power profiles improves deployment flexibility under data-center power constraints. It also suggests NVIDIA is increasingly factoring real-world physical infrastructure limits into product specifications.
2. VR200 NVL72 upgrades the GPU thermal solution for both MaxQ and MaxP.
? Both use a micro-channel cold plate (MCCP) paired with a gold-plated lid.
? The market has been expecting a move to a micro-channel lid (MCL) once GPU TGP reaches 2.3 kW; however, MCL mass production is unlikely before 2H27.
3. Benefiting from upgrades across the GPU, HBM, and NVLink Switch, VR200 NVL72 delivers roughly 3.5×/5× the training/inference AI computing power of GB300 NVL72, driving a sharp increase in rack-level power demand.
? VR200 NVL72 upgrades power shelves to 3*3U 110 kW (6*18.3 kW PSUs), vs. the most common GB300 NVL72 configuration of 8*1U 33 kW shelves (6*5.5 kW PSUs).
? VR200 NVL72 power shelves adopt an 3+1 redundancy design.
? VR200 NVL72 raises the power whip rating to 100A (vs. 60A for GB300 NVL72), tightening data-center power infrastructure requirements (e.g., busway and tap-off boxes).
4. VR200 NVL72 relies even more heavily on liquid cooling.
? Both the compute and NVSwitch trays adopt a fanless design.
? Rack technology cooling system flow (TCS flow) nearly doubles (~+100%) vs. GB300 NVL72, benefiting spec and/or count upgrades for CDUs, manifolds, cold plates, and quick disconnects (QDs).
? Rack airflow requirements fall by roughly 80% (in CFM) vs. GB300 NVL72.
5. VR200 NVL72 compute tray adopts a midplane for the first time, enabling a truly cableless design.
? Key midplane specs include 44 layers (22+22), M9 CCL (EM896K3), and an approximate size of 420×60 mm.
? Jensen noted that this design can reduce compute-tray assembly time from roughly two hours to about five minutes (vs. GB300 NVL72).
6. Rubin CoWoS are estimated at 300–350k wafers in 2026, with pilot production expected in early 1Q26 and mass production by late 2Q26. VR200 NVL72 rack assembly is expected to enter mass production by the end of 3Q26; factoring in yield ramp, 2H26 rack shipments are estimated at ~5,000–7,000 units.
7. Once rack power approaches ~200 kW, 54V distribution begins to face material constraints in space (busbars/cabling) and conversion efficiency. As a result, VR200 NVL72 can be viewed as the final generation of the Oberon rack architecture. To address the rising rack-power requirements driven by continued AI compute scaling, NVIDIA is expected to transition to the next-generation Kyber rack design, which supports 800V HVDC.
