Latest / DX Today | No-Hype Podcast & News About AI & DX / AI News Today: OpenAI Races to Raise $30B Before Anthropic IPO

Transcript

AI News Today: OpenAI Races to Raise $30B Before Anthropic IPO · DX Today | No-Hype Podcast & News About AI & DX

• 0:00
  SPEAKER_04: It's Tuesday, October the 20th, 2026.
• 0:03
  SPEAKER_04: You're listening to the DX Today AI Daily Brief.
• 0:07
  SPEAKER_04: Today, OpenAI races to raise$30 billion before its biggest rival goes public.
• 0:14
  SPEAKER_04: Anthropic weighs a second giant compute deal with NScale, and Chip Challenger 10 Storant lands$1.5 billion to take on Nvidia.
• 0:24
  SPEAKER_04: Let's get into it.
• 0:26
  SPEAKER_00: OpenAI is moving fast to lock in new money before rival Anthropic reaches the public markets.
• 0:31
  SPEAKER_00: Reuters reported on Monday, citing people familiar with the matter, that the ChatGPT maker is seeking about$30 billion in fresh investor commitments.
• 0:40
  SPEAKER_00: The round could value OpenAI at more than$1.4 trillion.
• 0:45
  SPEAKER_00: Existing backers SoftBank and an Abu Dhabi investment group are among those in talks, according to the report.
• 0:51
  SPEAKER_00: The timing matters.
• 0:52
  SPEAKER_00: Anthropic is expected to debut on the stock market in mid-November, and both companies are competing for the same pool of capital from sovereign wealth funds and large institutions.
• 1:02
  SPEAKER_00: The terms are not final, and the valuation could still change.
• 1:06
  SPEAKER_00: But the push shows how the race for AI funding is now shaped by the calendar of a rival's listing.
• 1:12
  SPEAKER_04: Now to the cloud deals.
• 1:14
  SPEAKER_05: Anthropic may be about to make another enormous bet on computing power.
• 1:19
  SPEAKER_05: Bloomberg reported that the British AI cloud company NScale is in talks to sell Anthropic compute capacity in a deal that could be worth about$45 billion.
• 1:31
  SPEAKER_05: It would be the second agreement between the two companies, and it would involve a second facility at NScale's campus in West Virginia.
• 1:40
  SPEAKER_05: NScale is backed by Nvidia, and it has grown quickly by building data centers designed for large AI workloads.
• 1:49
  SPEAKER_05: People familiar with the talks cautioned that nothing has been signed and the terms could change.
• 1:55
  SPEAKER_05: If it goes ahead, the deal would add to a wave of long-term capacity contracts as Frontier Labs try to guarantee the chips and power they need for the next generation of models.
• 2:07
  SPEAKER_03: Next, the chip money.
• 2:09
  SPEAKER_03: Tin Storent, the chip company led by veteran processor designer Jim Keller, has closed a$1.5 billion series E-round.
• 2:18
  SPEAKER_03: The financing values the company at$6.2 billion after the investment, up from$2.6 billion in its previous round.
• 2:26
  SPEAKER_03: The round was co-led by Trace Capital and Innovation X, with participation from Fidelity, Arkinvest, and T-Row Price, among others.
• 2:36
  SPEAKER_03: TanStorant says the money will support deployments of its Galaxy servers and further development of its AI and processor designs.
• 2:44
  SPEAKER_03: The company has positioned itself as an open alternative to NVIDIA, licensing its chip designs and building systems on an open instruction set.
• 2:53
  SPEAKER_03: The size of the round shows that investors still want options beyond the dominant suppliers of AI hardware.
• 3:00
  SPEAKER_03: Turning to security.
• 3:01
  SPEAKER_01: Microsoft says companies need to treat their AI agents like a new kind of insider.
• 3:06
  SPEAKER_01: On Monday, the company announced new capabilities for its project perception security system, including what it calls guardian agents.
• 3:15
  SPEAKER_01: These agents continuously watch other enterprise agents, assess their risk, and can recommend or apply containment when an agent starts to drift from the task it was given.
• 3:25
  SPEAKER_01: Microsoft also described a dynamic permissions approach, where an agent begins with no access and receives only the permissions it needs for approved work.
• 3:35
  SPEAKER_01: The risk score for each agent is updated using signals such as what data it can reach and how its behavior changes over time.
• 3:42
  SPEAKER_01: The features are part of Microsoft's top-tier workplace bundle.
• 3:46
  SPEAKER_01: The message is clear.
• 3:48
  SPEAKER_01: As agents spread, security teams need to supervise them the way they supervise people.
• 3:53
  SPEAKER_04: Now, the software battle.
• 3:55
  SPEAKER_02: Amazon is taking direct aim at one of Nvidia's strongest advantages, its software.
• 4:00
  SPEAKER_02: The company released Torch Neuron, a tool that lets developers run existing PyTorch code on Amazon's own training chips with a single configuration change and no rewrite.
• 4:12
  SPEAKER_02: PyTorch is the most widely used framework for building AI models, and much of that code has long been tied to Nvidia's CUDA software.
• 4:20
  SPEAKER_02: Amazon says Torch Neuron will be released under an open source license at the PyTorch Conference North America, which begins today.
• 4:28
  SPEAKER_02: The goal is to remove the friction that keeps customers from trying other hardware.
• 4:33
  SPEAKER_02: If developers can move their workloads without rewriting them, Trainium becomes a far easier choice inside Amazon Web Services.
• 4:41
  SPEAKER_02: Investors took notice, and Amazon shares rose sharply on Monday.
• 4:46
  SPEAKER_04: Over to Europe.
• 4:47
  SPEAKER_00: Nvidia is deepening its ties with Germany.
• 4:49
  SPEAKER_00: At a Berlin investment summit on Monday, the chipmaker said it plans to invest in four German startups, Agile Robots, Avelios Medical, Morello, and Reflex Aerospace.
• 5:01
  SPEAKER_00: The investments are part of NVIDIA's commitment of nearly$10 billion across Europe.
• 5:06
  SPEAKER_00: Although the individual amounts were not disclosed, the German government also signed a joint declaration of intent with NVIDIA to advance industrial AI, covering areas such as manufacturing and engineering.
• 5:19
  SPEAKER_00: Chancellor Friedrich Merz used the event to urge German businesses to embrace artificial intelligence and open up more of their data to it.
• 5:27
  SPEAKER_00: For Germany, the goal is to turn its industrial strength into an advantage in the AI era.
• 5:32
  SPEAKER_00: For NVIDIA, it is another step in building sovereign AI partnerships country by country.
• 5:37
  SPEAKER_05: In enterprise services, Cognizant and Google Cloud have expanded their partnership to push more businesses from AI experiments into real deployments.
• 5:48
  SPEAKER_05: The companies announced a new Gemini Frontier Capability Group backed by more than 18,000 Cognizant associates.
• 5:57
  SPEAKER_05: The group pairs Google's enterprise AI platform with Cognizant's industry and engineering expertise, and it builds on a dedicated Gemini practice that Cognizant launched in April.
• 6:10
  SPEAKER_05: The stated aim is to bridge the path from pilot to production, helping clients deploy agentic systems safely and at scale.
• 6:21
  SPEAKER_05: For Cognizant, the move is part of a broader shift by large services firms to retrain staff around AI building rather than traditional outsourcing.
• 6:32
  SPEAKER_05: For Google Cloud, it adds a large delivery force to help close enterprise yields against rivals, Microsoft and Amazon.
• 6:42
  SPEAKER_03: Now from China.
• 6:43
  SPEAKER_03: China's Minimax has released a faster and cheaper model built for AI agents.
• 6:48
  SPEAKER_03: The company launched M3.1 Flash, a successor to the M3 model it introduced in June.
• 6:55
  SPEAKER_03: Minimax says the new model cuts the cost of agent tasks by 70%, compared with its predecessor, and runs nearly three times faster.
• 7:04
  SPEAKER_03: It is aimed at complex enterprise workflows, software development, and computer use tasks.
• 7:10
  SPEAKER_03: Those figures are the company's own claims, and early independent testing has flagged a relatively high rate of hallucinations.
• 7:18
  SPEAKER_03: The launch comes just days after DeepSeek shipped an updated flash model of its own.
• 7:23
  SPEAKER_03: Together, the releases show Chinese developers competing hard on price and speed, as the market for agents that run many steps and consume large volumes of tokens keeps growing.
• 7:34
  SPEAKER_03: A different kind of model.
• 7:36
  SPEAKER_01: Inception, the startup behind commercial diffusion language models, has released Mercury 3.
• 7:42
  SPEAKER_01: Most large language models write text one word at a time.
• 7:45
  SPEAKER_01: Diffusion models instead refine a whole response in parallel, which can make them much faster.
• 7:51
  SPEAKER_01: Inception says Mercury 3 is its most capable model yet and sets a new frontier for the trade-off between speed and intelligence.
• 7:59
  SPEAKER_01: It is the company's fourth release since its first Mercury model in June of last year.
• 8:04
  SPEAKER_01: And Inception says each version has added capability without giving up speed.
• 8:08
  SPEAKER_01: The model is available now through the Inception API and partner platforms.
• 8:13
  SPEAKER_01: The performance claims come from the company itself.
• 8:16
  SPEAKER_01: Still, the launch is a sign that alternatives to the standard model architecture are moving from research into products that developers can use today.
• 8:24
  SPEAKER_02: In startup funding, a young AI startup called Town has become the latest unicorn.
• 8:30
  SPEAKER_02: The company raised$108 million in a Series B round at a$1 billion valuation, just four months after its public launch in June.
• 8:40
  SPEAKER_02: Town builds AI agents designed to get work done by collaborating with other agents on behalf of businesses.
• 8:47
  SPEAKER_02: The round was led by Forerunner Ventures and Menlo Ventures, with participation from Andreessen Horowitz, Serena Ventures, and First Round Capital.
• 8:56
  SPEAKER_02: That brings town's total funding to$160 million, following a$55 million Series A that closed in June.
• 9:06
  SPEAKER_02: The company says it reached seven figures in annual recurring revenue within weeks of launch.
• 9:11
  SPEAKER_02: The speed of the raise reflects how aggressively investors are still backing agent startups that show early revenue.
• 9:18
  SPEAKER_04: Turning to robotics.
• 9:19
  SPEAKER_00: Dream Labs, a San Francisco startup founded by former NVIDIA robotics researchers, has raised$90 million in seed funding.
• 9:27
  SPEAKER_00: The round was led by B Capital with participation from Kindred Ventures, Unusual Ventures, SV, Angel, and Wing VC, among others.
• 9:36
  SPEAKER_00: Dream Labs is building a general purpose brain for everyday robots, software that lets machines from different manufacturers perform tasks without extensive custom programming.
• 9:47
  SPEAKER_00: Alongside the funding, the company launched its first robot model, Dream One and Dream API, a developer platform that turns camera input and plain language instructions into robot actions.
• 10:00
  SPEAKER_00: Chief Executive Joel Jang co-founded the company.
• 10:04
  SPEAKER_00: A seed round of this size is unusual, and it shows how much capital is flowing toward foundation models for robotics as investors look for the next platform beyond chatbots.
• 10:14
  SPEAKER_05: And finally, data for agents.
• 10:27
  SPEAKER_05: Boomi says the deal adds data quality, observability, governance, and discovery tools to its platform, with the goal of giving companies trusted data for their AI systems.
• 10:40
  SPEAKER_05: Israeli business outlet CTEC reported the price at roughly$40 to$45 million, though Boomi did not disclose terms.
• 10:49
  SPEAKER_05: The reasoning is simple.
• 10:51
  SPEAKER_05: AI agents are only as reliable as the data they act on, and broken pipelines or stale tables can lead them to confident but wrong answers.
• 11:01
  SPEAKER_05: Elementary will continue to operate as a standalone product for now.
• 11:06
  SPEAKER_05: The deal is part of a broader trend of data and integration vendors buying tools that make enterprise data ready for agents.
• 11:15
  SPEAKER_04: That's your briefing for Tuesday, October the 20th, 2026.
• 11:20
  SPEAKER_04: For DX today, stay curious.