Rumors of NVIDIA RU chips are seen as a 6G architectural evolution rather than a negative for equipment vendors

Overview

This is a JPMorgan Tech Snippets research note primarily discussing whether NVIDIA will extend its AI-RAN strategy from central/distributed unit (CU/DU) compute to mobile network radio units (RU). The report believes that as 6G and Massive MIMO push more low-PHY processing toward RUs, GPU or programmable silicon solutions become more attractive relative to fixed-function ASICs; however, this does not directly constitute a negative for Nokia or Ericsson.

Core views

The core view of the report is that NVIDIA's entry into RU is not surprising and should not be seen as an effective bearish thesis for Nokia or Ericsson. Nokia is already cooperating with NVIDIA at the baseband CU/DU level, so if it decides in the future to fully shift to a standardized semiconductor platform, it may further adopt NVIDIA solutions in RU. Ericsson is currently not considering abandoning ASICs because it values the performance advantages of custom chips, but it may still reassess in the future based on the relative advantages of NVIDIA's solution versus its own ASIC roadmap. The report also notes that 6G is still at least three years away, and the market's attribution of Nokia's and Ericsson's stock declines to NVIDIA-related news may be post hoc reasoning.

Analysis framework

The report uses an event commentary and technology path comparison approach: it first reviews rumors about NVIDIA RU chips and the low-PHY compute demand brought by Massive MIMO, then compares the trade-offs between standard GPU/AI accelerator platforms and custom ASICs in flexibility, performance, power consumption, and upgrade cadence, and finally assesses the potential impact on Nokia, Ericsson, and NVIDIA ecosystem partners.

Methodology notes

• Industry technology analysisAI-RAN and RU architecture evolution

  Migration of low-PHY compute toward RU

  The report believes that the expansion of Massive MIMO scale significantly increases low-PHY processing needs such as beamforming, creating a technical window for programmable GPU solutions to enter RU.

• Competitive path comparisonComparison of standard platforms and custom ASICs

  Trade-off between flexibility and performance

  Standard semiconductor platforms are conducive to decoupled software-hardware upgrades and lower hardware upgrade costs, but custom ASICs usually provide stronger performance advantages, requiring operators to choose between flexibility and performance.

• Event impact assessmentTesting the hypothesis of a negative for equipment vendors

  Whether NVIDIA's RU entry disrupts Nokia/Ericsson

  The report's conclusion is no; both equipment vendors can still choose or adjust their product roadmaps, and there remains a long time window before 6G commercialization.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• NVIDIA CORP (NVDA)

  Potential supplier of RU GPU solutions and AI-RAN platforms.

  Strengths

  It has the GPU, CUDA, and AI accelerator ecosystem, and may integrate AI applications and radio workloads on the same platform.

  Weaknesses

  In RU scenarios it must meet strict power, thermal, and performance requirements, and needs to prove it can replace custom ASICs.

  Comparison

  Compared with fixed-function ASICs, NVIDIA's solution is more programmable and flexible; however, ASICs may still have the upper hand in performance and targeted optimization.

  Risks

  Excessive power consumption, a performance gap versus dedicated ASICs, and uncertainty around 6G standards and equipment vendor adoption timing.

• Nokia (NOKIA.HE/NOK)

  An equipment vendor that may adopt NVIDIA standard platforms in CU/DU and RU.

  Strengths

  It has stated that it will adopt NVIDIA AI accelerators in 6G baseband CU/DU, and already has a cooperation foundation around NVIDIA chips and CUDA.

  Weaknesses

  If it fully shifts to standardized semiconductors, hardware differentiation would decline, and more differentiation would need to come from software.

  Comparison

  Compared with Ericsson, Nokia may have a higher probability of adopting NVIDIA RU solutions and may implement them faster.

  Risks

  Operators may prefer ASIC performance more; platform switching would bring product architecture and customer acceptance risks.

• Ericsson (ERICb.ST/ERIC)

  An equipment vendor that adheres to ASIC performance advantages but needs to assess standard platform solutions.

  Strengths

  It has an in-house ASIC roadmap and performance advantages, and is not currently considering abandoning ASICs.

  Weaknesses

  It lacks experience with NVIDIA architecture, and if it shifts to NVIDIA platforms in the future, the required architectural changes may be larger and take longer.

  Comparison

  Compared with Nokia, Ericsson faces a higher hurdle in switching to NVIDIA solutions, but it does not lack optionality.

  Risks

  If standard platforms gain clear cost or ecosystem advantages in the 6G era, Ericsson may face pressure to adjust its roadmap.

• Marvell

  An NVIDIA ecosystem partner that plans to enable its Octeon base station processor to work directly with NVIDIA GPUs.

  Strengths

  Its base station processor can be combined with GPUs to support co-platform operation of radio workloads and AI applications.

  Weaknesses

  The report does not provide Marvell-related ratings or quantified financial impact.

  Comparison

  Marvell is more of a connectivity and processor partner within the NVIDIA AI-RAN ecosystem rather than an RU equipment vendor.

  Risks

  Solution rollout depends on equipment vendor adoption, progress in 6G standards, and validation of power/performance.

Key data

• Report date2026-06-09The report disclosure completion time was 09 Jun 2026 06:06 PM BST, and the distribution time was 09 Jun 2026 06:08 PM BST.
• NVDA rating and priceOW / $208.64The company list shows NVIDIA Corporation with an OW rating, and the price is the closing price under the report's convention.
• Nokia rating and priceNOKIA.HE: €11.97 / OW；NOK：$14.59 / OWThe report believes that because Nokia already cooperates with NVIDIA at the CU/DU level, the probability of adopting NVIDIA RU solutions in the future is relatively higher.
• Ericsson rating and priceERICb.ST：Skr111.45 / N；ERIC：$12.52 / NThe report believes that although Ericsson is more committed to the ASIC path, NVIDIA's RU entry is not an effective bearish argument.
• MIMO compute expansionThe move from 4T/4R to 128T/128R brings about roughly 32x low-PHY compute demandThe report also mentions that the future ultra-MIMO concept may reach 1,024T/1,024R.
• RU energy consumption pressureAbout 90% of network energy consumption comes from RUPower consumption is one of the key risks facing NVIDIA-style GPU RU solutions.
• Ecosystem investmentMarvell $2bn；Nokia $1bnThe report mentions NVIDIA's recent investments in Marvell and Nokia, and says Marvell plans to enhance its Octeon base station processor to work directly with NVIDIA GPUs.
• Time window6G is still at least three years awayThe report believes short-term stock price fluctuations should not be simply attributed to NVIDIA RU news.

Impact & implications

For NVIDIA, RU chips represent a potential incremental scenario in which AI-RAN extends from baseband compute to the radio side, but it must prove that power consumption and performance can approach or exceed those of dedicated ASICs. For Nokia, existing collaboration experience with NVIDIA may reduce the migration cost of adopting standard platforms. For Ericsson, sticking with ASICs in the short term still has a performance rationale, but if standard platforms prove superior in software flexibility, integration of AI workloads, and cost, it may also be forced to reassess its roadmap. Overall, the report interprets this news as a 6G architecture choice issue rather than a direct negative catalyst for traditional equipment vendors.

Risks

• RUs account for about 90% of network energy consumption, so GPU-style solutions must solve power consumption and thermal issues.
• Custom ASICs may still outperform standard GPU or AI accelerator platforms.
• Operators may value performance more than the flexibility brought by standard hardware.
• 6G is still at least three years away, and there remains uncertainty around technology roadmaps, standards, and commercialization timing.
• The report does not provide an NVDA target price or detailed valuation model; valuation and company-specific risks require consulting JPMorgan's company research.

What to watch

• Whether Nokia expands from CU/DU to RU in adopting NVIDIA solutions.
• Whether Ericsson continues to stick with the ASIC path or reassesses standard platforms before 6G.
• Whether NVIDIA can launch an RU design below the 100W class, with power characteristics close to automotive embedded GPUs.
• Progress in the coordinated solution between Marvell Octeon base station processors and NVIDIA GPUs.
• Whether Massive MIMO and ultra-MIMO demand drive more low-PHY processing toward RU.
• Whether the market continues to use NVIDIA RU news to explain the stock price movements of Nokia and Ericsson.