SoftBank to Buy ABB Robotics for $5.4B

The deal at a glance

• SoftBank Group will acquire ABB’s robotics division for $5.4 billion.
• The transaction ends ABB’s plan to list the unit separately.
• Both companies frame the deal as a catalyst for AI-led robotics.

Transaction overview

SoftBank Group has agreed to purchase ABB’s robotics division for $5.4 billion, expanding its push into AI-powered machines and ending ABB’s plan to list the unit separately.

ABB said it will deploy the proceeds according to its established capital allocation framework, without detailing specific uses. The company had previously explored carving out the robotics business, but will no longer pursue a separate listing following the deal.

Strategic vision

ABB CEO Morten Wierod said both companies see the industry entering a new era of AI-led robotics and believe combining ABB Robotics with SoftBank’s portfolio positions them to help shape that shift.

“Physical AI” is central to SoftBank’s next focus, bringing together leading technology and talent to accelerate the fusion of advanced AI and robotics.

SoftBank founder and CEO Masayoshi Son has characterized artificial super intelligence (ASI) as a stage beyond artificial general intelligence (AGI), potentially thousands of times more capable than human intelligence and achievable within the next decade. He envisions ASI-enabled robots taking on a wide range of physical tasks, from manufacturing and transportation to construction and household work.

SoftBank’s broader AI push

The acquisition aligns with SoftBank’s strategy to be a central player in the AI ecosystem. In recent years, the company has built stakes in chip designers Arm and Ampere and invested in OpenAI. SoftBank also previously ventured into humanoid robots with its Pepper model, whose production reportedly ended in 2020.

Integration outlook and industry impact

Few operational details were disclosed about how SoftBank plans to integrate ABB Robotics or prioritize product lines. However, the move underscores rising interest in embedding AI more deeply into industrial automation and service robotics.

By combining SoftBank’s AI ambitions with ABB’s robotics expertise, the deal aims to accelerate the development of intelligent machines while providing ABB with capital flexibility. If the partnership delivers on its vision, it could help set the pace for the next phase of AI-driven robotics across factories, logistics, and everyday environments.

Key takeaways

• $5.4B acquisition halts ABB’s planned spin-off of its robotics unit.
• SoftBank targets “physical AI” and long-term ASI-driven robotics.
• Partnership could accelerate intelligent automation across industries.

EU unveils $1.1B push for AI sovereignty

EU earmarks $1.1B to speed up homegrown AI

The European Commission’s $1.1B Apply AI Strategy funds skills, commercialization, and alliances to boost EU AI; experts say more investment is needed.

The European Commission has unveiled a $1.1 billion package to accelerate the use of artificial intelligence across Europe, aiming to strengthen technological sovereignty and close competitive gaps.

What the Apply AI Strategy includes

• Make Europe’s workforce “AI‑ready”: expand training and upskilling to bring AI skills into mainstream roles.
• Shorten the lab‑to‑market path: help researchers and startups commercialize more quickly.
• Connect key players: link academia, industry, and the public sector to accelerate adoption in priority sectors.

Two pillars to support adoption

1) Apply AI Alliance

A forum uniting innovators from government, academia, and civil society to share best practices, coordinate projects, and reduce fragmentation across the EU.

2) AI Act Service Desk

A guidance hub to help organizations implement the EU’s AI Act, which took effect in August 2024. The desk aims to clarify compliance requirements and lower the cost of adoption for SMEs and public bodies.

Funding will be channelled through existing EU programs, including Horizon Europe and DigitalEurope.

Competitive context: the U.K. surge

Brussels’ push comes amid concerns that Europe is lagging in AI. Outside the EU, the U.K. has drawn significant private investment: Nvidia announced plans on Oct. 7 to invest $3.5 billion in the British AI sector, supporting startups including PolyAI and Basecamp.

Nvidia‑backed cloud GPU provider Nscale recently raised $1.1 billion, and Nvidia and Nscale outlined a $13 billion commitment to U.K. AI infrastructure, signalling intense competition for capital, talent, and compute.

What analysts are saying

Analysts say the EU program is a necessary start, but not sufficient on its own. Torsten Volk of Omdia called it a catalyst that will require coordinated support from member states to havea  real impact.

While $1.1 billion can help startups overcome upfront capital hurdles and clear early bottlenecks, much larger investments in talent, data centers, energy, and core infrastructure are needed to scale successful generative AI applications.

Why it matters

• Signals a clear intent to accelerate adoption and reduce EU fragmentation.
• Could unlock stalled public and private projects by guiding the AI Act.
• Creates a cross‑sector alliance to speed pilots and commercialization in strategic industries.

The bottom line

The Apply AI Strategy signals intent and could unlock stalled projects, but its long‑term effect will depend on follow‑through by EU countries and additional funding well beyond the initial package.

Oracle, SoftBank to launch Japan sovereign cloud, AI

Oracle and SoftBank are expanding their collaboration to deliver sovereign cloud and AI services designed for Japanese enterprises, addressing rising requirements for data residency, security, and regulatory compliance.

SoftBank’s Cloud PF Type A on Oracle Alloy

SoftBank’s upcoming Cloud PF Type A, its proprietary cloud and AI platform, will be built on Oracle Alloy for the Japan market. The companies plan a phased rollout of AI capabilities over the next several years to support the country’s digital transformation. Oracle Alloy will provide SoftBank access to more than 200 cloud and AI services, operated within SoftBank’s domestic data centers.

Deployment Footprint and Timeline

SoftBank intends to deploy the platform across its eastern and western Japan facilities, enabling customers to run sensitive workloads while maintaining strict control over data location and governance.

Security and Key Management

To strengthen protections, SoftBank will pair an Oracle service for centrally managing encryption keys with its own key management technology. The combined approach is designed to enhance customer control over cryptographic keys and reinforce defences against unauthorized data access.

Executive Perspectives

Broader Collaboration

The Japan initiative builds on the growing Oracle–SoftBank relationship. In the U.S., the companies are working with OpenAI on the $500 billion Stargate program announced in January to establish large-scale AI infrastructure. In late September, the partners outlined plans for five new AI data centers under that effort.

What does it mean for the Japanese enterprises?

• Sovereign-by-design operations with data residency and governance controls
• Access to 200+ Oracle cloud and AI services through SoftBank’s domestic platform
• Enhanced key management to reduce the risk of unauthorized data access
• Phased AI rollout supported by high-performance GPUs for modern workloads

Nvidia, Fujitsu team on Japan AI, robotics infrastructure

What was announced

Announced in Tokyo, the initiative brings together Nvidia’s GPUs with Fujitsu’s CPUs and orchestration technologies to create end-to-end platforms for enterprises. The partners plan to co-develop “self‑evolving” AI agents tailored to specific sectors, initially manufacturing, healthcare, and robotics, so organizations can deploy systems that learn and improve over time.

Executive perspective

Executives framed the effort as groundwork for a new wave of industrial automation. Nvidia CEO Jensen Huang said the AI industrial revolution is underway and emphasized the need to build the supporting infrastructure in Japan and globally. Fujitsu CEO Takahito Tokita said the companies will deliver full‑stack AI capabilities starting in areas where Japan leads, and intend to broaden the partnership into high‑performance computing and quantum technologies to meet growing demand.

Timeline, scope, and potential partners

While neither firm disclosed project budgets or detailed timelines, they set a goal of establishing a national AI hub in Japan by 2030. Robotics is expected to be a major pillar of the plan, and the companies indicated that a potential collaboration with Japanese robot maker Yaskawa Electric is under consideration.

Why does it matter?

Fujitsu positioned the announcement against surging interest in generative AI, where high costs and technical complexity have tended to favour large enterprises. The partners said their joint platform aims to lower those barriers and widen access to advanced AI tools.

The news comes amid exceptional demand for Nvidia’s chips, which power many state‑of‑the‑art AI systems. Nvidia’s market capitalization surpassed $4.5 trillion this week, marking a new high for the company.

What to watch next

What comes next will be concrete project details, sites, investments, and rollout milestones, as the partners work toward the 2030 target. If successful, the collaboration could give Japanese industry a stronger local footing for AI development and accelerate the deployment of intelligent robots on factory floors and in healthcare settings.

a16z: Consumer AI Is Moving Into Enterprise

What the new a16z/Mercury report says

Consumer-grade AI is quietly becoming a gateway into the enterprise, according to a new report from venture capital firm Andreessen Horowitz (a16z) compiled with fintech company Mercury.

How were the rankings built?

Drawing on anonymized transaction data from Mercury’s platform, a16z assembled a Top 50 list of AI‑native application layer companies to show where startups are actually spending on AI products. The analysis extends a16z’s prior Top 100 Gen AI Consumer Apps report and focuses on real usage in products and workflows rather than hype.

Top AI apps by startup spend

OpenAI, maker of ChatGPT, leads the rankings. Other large language model assistants also placed highly, with Anthropic at No. 2 and Perplexity at No. 12. Coding tool Replit took the No. 3 spot, while creative suite Freepik ranked No. 4. Meeting support tools featured in the “horizontal” productivity category included Fyxer (No. 7), Happyscribe (No. 36) and Plaud (No. 38).

• LLM assistants: OpenAI (No. 1), Anthropic (No. 2), Perplexity (No. 12)
• Developer tools: Replit (No. 3)
• Creative suite: Freepik (No. 4)
• Meeting support: Fyxer (No. 7), Happyscribe (No. 36), Plaud (No. 38)
• Vertical tools: Lorikeet (No. 8), Micro1 (No. 9), Instantly (No. 13)

Horizontal vs. vertical tools

a16z groups these offerings into two broad types. Horizontal tools aim to raise overall productivity across roles, such as general LLM assistants, creative apps and meeting aids.

“Vertical” tools target specific functions and, for now, typically augment workers on repetitive tasks rather than replace them. Popular vertical categories include customer service, sales and recruiting, where Lorikeet, Micro1, and Instantly stand out.

Bottom‑up adoption inside enterprises

The report highlights a notable adoption pattern: consumer-first AI tools are increasingly spreading inside companies. Roughly 70% of the listed products can be adopted by individuals and brought into teams because they don’t require an enterprise license. With organizations under pressure to boost employee efficiency, these tools are being pulled into enterprise environments faster than in prior software waves.

Why does it matter?

Taken together, the spending signals suggest where AI is delivering practical value today, across general-purpose assistants, developer tools and targeted applications that slot into everyday workflows.

Looking ahead

a16z expects this bottom‑up motion to persist, predicting that many of the next enterprise AI successes will begin life as consumer products. The firm plans to keep updating the online list to reflect shifting spend and usage.

Meta signs $14.2B AI compute deal with CoreWeave

Meta taps CoreWeave in $14.2B AI compute pact

CoreWeave has secured a multiyear, $14.2 billion agreement to provide Meta Platforms with cloud-based AI compute, extending into the next decade as the tech industry races to lock down processing power. The contract runs through 2031 with an option to renew for an additional year.

Key points

• Long-term pact runs through 2031, with a one-year renewal option.
• Deal reflects the scramble to secure GPUs for training and deploying generative and agentic AI.
• Analysts expect AI compute demand to stay strong into the 2030s despite bubble concerns.

Deal context and market momentum

The deal arrives amid a broader surge in AI infrastructure spending. It follows reports of a far larger, $100 billion arrangement between OpenAI and Oracle to build massive data centers for generative AI, and comes alongside other headline investments, including Nvidia’s reported $100 billion commitment to OpenAI and $5 billion to U.S. chipmaker Intel. The momentum reflects escalating demand for compute to train and run generative and agentic AI models, as well as emerging use cases like humanoid robotics.

Why CoreWeave, and what it signals

Analysts say Meta’s move validates CoreWeave’s position. Nick Patience of the Futurum Group noted that Meta needs reliable access to compute and that turning to CoreWeave underscores confidence in the provider’s capacity well into 2031. CoreWeave sits among a newer cohort of AI-focused cloud compute companies, alongside Nscale, Lambda Labs, and Nebius, competing with hyperscalers such as AWS, Google, and Microsoft, which are simultaneously developing their own AI chips and infrastructure. Nvidia, which operates its own cloud platform, is also a major force in the AI compute market.

“Meta needs reliable access to compute, and turning to CoreWeave underscores confidence in its capacity well into 2031,” said Nick Patience of the Futurum Group.

What it means for Meta

CoreWeave’s customer roster has been anchored by Microsoft to date, and it also supplies OpenAI. For Meta, the agreement supports its expansive AI ambitions. The company has embedded generative AI across Facebook, Instagram, and WhatsApp and is a leading backer of open models through the Llama series.

While Meta is building proprietary AI data centers, Omdia analyst Torsten Volk observed that CoreWeave can provision GPUs at a larger scale, optimize utilization, and ensure power availability, even during peak demand, outstripping Meta’s in-house capacity. He added that Meta’s priority is advancing AI products rather than operating a GPU hyperscale cloud.

Is there a risk of overbuild?

The scale of current projects has sparked debate about whether today’s AI build-out could overshoot demand. In addition to the OpenAI–Oracle initiative, the sprawling Stargate data center effort involving OpenAI, Oracle, SoftBank, and others has fueled questions about potential overcapacity.

Patience, however, expects AI compute needs to remain strong into the 2030s. He acknowledged bubble-like elements in the market but argued it’s unlikely these facilities will be idled mid-decade, given ongoing growth in AI processing requirements.

The bottom line

By tying up long-term compute with CoreWeave, Meta signals it intends to sustain, and expand, its AI efforts for years to come. The pact also highlights how partnerships between AI developers and specialized cloud providers are becoming central to securing the GPU resources that underpin the next wave of AI applications.

Cerebras Raises $1.1B at $8.1B Valuation in Series G

Overview

Cerebras Systems has secured $1.1 billion in new funding, lifting the Silicon Valley AI chip maker’s valuation to $8.1 billion and postponing its move to the public markets for now.

Founded in 2015, Cerebras positions itself as a challenger to Nvidia with purpose-built chips, systems, and cloud services for AI workloads. The company filed for an IPO in September 2024, but said the fresh capital will allow it to accelerate growth privately.

Round Details and Investors

The oversubscribed Series G round was led by Fidelity Management & Research Company. Returning investors included Altimeter, Alpha Wave, and Benchmark, alongside Tiger Global, Valor Equity Partners, and 1789 Capital.

How Cerebras Plans to Use the Capital

• Advance AI processor design, packaging, and system engineering.
• Build next-generation AI supercomputers.
• Expand U.S. manufacturing footprint.
• Add data center capacity to meet surging demand.

Inference Services and Performance

A key driver is the company’s AI inference services, launched in August 2024. Cerebras cites third-party research from Artificial Analysis indicating its system delivers inference speeds up to 20 times faster than Nvidia GPUs.

Artificial Analysis CEO Micah Hill said the firm has tested providers across hundreds of models and found Cerebras consistently on top.

Data Center Expansion

To support demand, Cerebras in March announced six additional inference data centers slated to open across North America and Europe in 2025, with locations including Minneapolis, Oklahoma City, and Montreal.

Partners, Customers, and Ecosystem

The company pointed to a growing roster of partners and customers such as AWS, Meta, IBM, GlaxoSmithKline, the U.S. Department of Energy, and the U.S. Department of Defense.

On the developer platform Hugging Face, Cerebras says it is the leading inference provider, now handling more than five million requests each month.

Financing Timeline and IPO Plans

With this round, Cerebras’ total funding raised over the past decade approaches $2 billion. Co-founder and CEO Andrew Feldman said long-time backers see a generational opportunity in AI and have chosen to double down on the business.

Feldman told Reuters that Cerebras still intends to go public. He said the timeline was affected by a national security review of a $335 million investment tied to Abu Dhabi, which he said has been cleared by the Trump administration. He added that raising additional capital late in the IPO journey is common among high-growth companies.

The Bigger Picture

The capital infusion underscores investor appetite for alternatives to Nvidia and intensifies the race to deliver faster, more efficient compute for AI applications.

NIST flags security risks in DeepSeek AI

NIST says China’s DeepSeek trails U.S. rivals in cybersecurity and reasoning, shows censorship and data‑sharing risks. Experts urge cautious, controlled use.

Overview

A U.S. government assessment has raised red flags about Chinese AI vendor DeepSeek, finding its models underperform U.S. rivals on key security and reasoning measures and exhibit politically aligned censorship and data‑sharing behaviors.

What NIST Found

Released Sept. 30 by the National Institute of Standards and Technology’s Center for AI Standards and Innovation (CAISI), the report says DeepSeek systems lag OpenAI’s GPT‑5 and Anthropic’s Claude Opus 4 across several benchmarks. NIST found the models are more vulnerable to “agent hijacking” attacks, attempts to capture credentials or redirect actions, and more likely to follow malicious instructions.

• More vulnerable to “agent hijacking” attempts to capture credentials or redirect actions.
• More likely to follow malicious instructions.
• Political bias consistent with positions favored by China’s government, including claims that Taiwan is part of China.
• Models share user data with outside parties, naming ByteDance among them.

Policy Context

CAISI said it produced the evaluation in response to President Donald Trump’s AI Action Plan, which asked the agency to assess models from China.

Market Background

The report lands about a year after DeepSeek‑R1 briefly stunned the market by approaching Western performance while using far less compute and capital. Although DeepSeek’s open‑source releases never matched the popularity of ChatGPT, they drew a global user base and influenced rivals, reducing momentum for some open‑weight alternatives like Meta’s Llama and nudging OpenAI to offer more open options earlier in the year.

Performance Highlights

Performance is not uniformly weak, NIST noted. On science Q&A and knowledge tests, DeepSeek models performed comparably to U.S. peers; DeepSeek V3.1 ranked near the top on scientific, mathematical and reasoning tasks.

But American models generally led in software engineering and cybersecurity, according to CAISI’s findings.

Expert Perspectives

Censorship and Model Values

Kashyap Kompella, CEO of RPA2AI Research, said the assessment shows how large language models tend to reflect the values and constraints of their creators. He argued censorship is structurally embedded in China‑based systems due to domestic rules and remains even when models are open‑sourced or locally hosted. He also noted that U.S. models are constrained by corporate guardrails and commercial considerations.

Skepticism on Earlier Claims

Some industry analysts questioned DeepSeek’s earlier performance claims. David Nicholson of Futurum Group said projections that enterprises would retool around custom GPU software stacks to replicate DeepSeek’s gains are unrealistic. The more pressing issue, he said, is how organizations can ensure security as they rely more on LLMs and generate fresh data continuously.

Enterprise Guidance

For companies that still want to test DeepSeek, Nicholson advised containing risk by accessing the models through enterprise‑grade platforms such as AWS Bedrock or Microsoft Azure and operating within tightly controlled environments.

• Access via vetted enterprise platforms (e.g., AWS Bedrock, Microsoft Azure).
• Operate within tightly controlled, monitored environments.
• Prioritize providers that meet security and regulatory expectations.

The Bottom Line

Taken together, NIST’s findings intensify scrutiny of DeepSeek at a time when global model ecosystems appear to be specializing, Chinese systems showing strength in scientific reasoning, and U.S. models leading in software and security.

For enterprise buyers, the takeaway is pragmatic: validate performance claims in real workloads, audit security and privacy behavior, and weigh geopolitical and compliance exposure before deployment.

Meta rolls out Business AI for SMBs, AR shopping features

Meta debuts Business AI for SMBs on its apps and websites, plus AR try-on and an ads assistant. Analysts see promise but warn of integration hurdles.

Overview

Meta has unveiled a new suite of AI features aimed at small and midsize businesses, led by a sales-focused agent called Business AI that works across Meta’s platforms and on company websites.

What Business AI does

Business AI is designed to boost conversions by handling customer inquiries, automating parts of the sales cycle, and simplifying AI adoption for smaller teams. Unlike tools confined to social apps, Meta says the agent can be embedded on a business’s own site to assist shoppers and respond to questions in real time.

• Handles customer questions and product guidance in real time
• Automates pieces of the sales process to drive conversions
• Embeddable on websites for on-site assistance, not just social
• Built to help smaller teams adopt AI without heavy lift

New shopping and advertising experiences

Meta also previewed new shopping and advertising features designed to enhance personalization and speed to creative.

• AR try-on: Customers can see how products look on them before buying.
• Custom calls to action: A capability in testing lets marketers upload their own images to generate customized CTAs within ads.

Support for advertisers: Meta AI business assistant

To support advertisers, the company introduced the Meta AI business assistant, a chat experience inside Ads Manager and Business Support Home. The assistant provides tailored, data-driven guidance and can help resolve account issues more efficiently, according to Meta.

The competitive backdrop

The move lands in a crowded market. Microsoft introduced a sales agent earlier this year, and Google continues to expand AI tools for businesses. Analysts expect most major vendors to launch similar agents as demand grows.

Analyst perspective

Bottom line

Meta’s new AI capabilities aim to streamline marketing and sales for SMBs, especially on Facebook and Instagram, while success off-platform will hinge on thoughtful integration and governance.