Figure Valuation Soars to 270 Billion RMB; Unitree Open-Sources 'Trump Card' Tech; Nvidia Invests 35.5 Billion RMB in Intel | Electric Wave Weekly Report

1. Figure completes over $1 billion in funding, post-money valuation reaches $39 billion

On September 16 local time, humanoid robot company Figure announced the completion of a Series C funding round exceeding $1 billion, post-money valuation reaching $39 billion (approximately RMB 270 billion), breaking the global valuation record for humanoid robot companies.

This round of financing was led by Parkway Venture Capital, with other investors including Nvidia, Brookfield Asset Management, Intel Capital, LG Technology Ventures, Salesforce, T-Mobile Ventures, and Qualcomm Ventures.

Figure stated that this round of financing will continue to support the company's development in three key areas: promoting the deployment of humanoid robots in home and commercial scenarios, building next-generation GPU infrastructure, and initiating advanced data collection work for Helix.

On September 17 and 18, following the announcement of its financing completion, Figure separately announced a partnership with alternative asset management firm Brookfield to create large-scale real-world home scenarios for training the Helix model; and launched the Go-Big project, aiming to build the world's largest and most diverse humanoid pre-training dataset.

**2. Unitree Opens Source of the UnifoLM-WMA-0 Architecture

On the evening of September 15, Unitree announced the open-sourcing of the UnifoLM-WMA-0 architecture, a world model-action architecture under the UnifoLM series designed for general-purpose robot learning.

The core component of the UnifoLM-WMA-0 architecture is a world model that can understand the physical laws governing robot-environment interactions. It features two core capabilities:

• Simulation engine, which operates as an interactive simulator to provide synthetic data for robot learning;
• Policy enhancement, which can be connected to an action head to further optimize decision-making performance by predicting future interactions with the physical world.

Furthermore, under the UnifoLM-WMA-0 architecture, the world model is not limited to a single mode; it supports two operational modes—decision mode and simulation mode—offering considerable flexibility.

In decision mode, the model can predict information regarding future physical interactions between the robot and its environment, assisting the policy in generating actions. This allows the robot to "think ahead" about the potential consequences of its actions during task execution, thereby making better decisions. In simulation mode, the model generates high-fidelity environmental feedback based on the robot's actions, effectively simulating an interaction environment that closely resembles reality.

According to information on Unitree's official website, UnifoLM has previously been integrated into the Unitree G1. Regarding this full open-source release of UnifoLM-WMA-0, Unitree Technology stated that it will continue to update the project in the future.

3. NVIDIA invests 35.5 billion RMB in Intel, joining forces to focus on chips

On the evening of September 18 Beijing time, Nvidia and Intel announced a partnership. Nvidia will invest $5 billion (approximately 35.5 billion yuan) to acquire shares in Intel, becoming a major shareholder.

The two companies will jointly develop multiple generations of customized data center and personal computing products. In the data center sector, Intel will build customized x86 CPUs for NVIDIA, which will be integrated into NVIDIA's AI infrastructure platform for market launch.

In the personal computing segment, Intel will research and launch system-on-chips (SoCs) with an x86 architecture that integrate NVIDIA RTX GPU chipsets. This new x86 RTX SoC will power various PC products requiring advanced integrated CPU and GPU solutions.

Notably, this investment agreement primarily focuses on cooperation between NVIDIA and Intel's product divisions rather than their foundry businesses, though they have not ruled out potential future collaboration in the semiconductor manufacturing sector.

4. Meta Unveils Next-Generation Smart Glasses 'Meta Ray-Ban Display'

On September 18, Meta launched its next-generation smart glasses, 'Meta Ray-Ban Display,' alongside the innovative interactive device 'Meta Neural Band.'

In terms of size, the frames are available in standard and large sizes, with black and sand color options.

The glasses weigh approximately 69 grams, offering 6 hours of battery life for mixed use, and up to 30 hours total when paired with the portable, foldable charging case.

Functionally, the Meta Ray-Ban Display features full-color, high-resolution lens displays that support viewing messages, photo previews, navigation, live captions, and real-time translation.

Each pair of glasses comes with the Meta Neural Band, which interprets natural signals from muscle activity to control the glasses' functions. Users can operate the experience via gestures without touching the glasses or their phones. The band is IPX7 waterproof and offers a single charge lasting up to 18 hours.

Priced at $799 (including the glasses and the band), the new Meta Ray-Ban Display will go on sale in the United States on September 30.