Latest / DX Today | No-Hype Podcast & News About AI & DX / AI News Today: Oracle Opens Its Cloud to Google's AI Chips

Transcript

AI News Today: Oracle Opens Its Cloud to Google's AI Chips · DX Today | No-Hype Podcast & News About AI & DX

• 0:00
  SPEAKER_05: It's Wednesday, October 28, 2026.
• 0:03
  SPEAKER_05: You're listening to the DX Today AI Daily Brief.
• 0:07
  SPEAKER_05: Today, Oracle opens its cloud to Google's AI chips.
• 0:11
  SPEAKER_05: SK Heinex posts a record quarter on surging memory demand, and Google Deep Mind unveils Genie 4, a world model that now comes with sound.
• 0:22
  SPEAKER_05: Let's get into it.
• 0:23
  SPEAKER_05: First, the cloud chip race.
• 0:26
  SPEAKER_00: Oracle is widening the menu of chips inside its cloud.
• 0:29
  SPEAKER_00: At its AI World Conference in Las Vegas on Tuesday, the company said Oracle Cloud Infrastructure will offer Google's tensor processing units starting next year, ramping over several years.
• 0:40
  SPEAKER_00: It is the first time Google's custom AI chips will be available on a large cloud outside Google's own.
• 0:46
  SPEAKER_00: Oracle also plans to deploy inference systems from Cerebrus in the first half of next year and to add hardware from Salmanova.
• 0:54
  SPEAKER_00: At the same time, Oracle and NVIDIA said they are expanding NVIDIA's Vera Rubin systems on Oracle's cloud to multi-gigawatt scale.
• 1:03
  SPEAKER_00: So this is not a retreat from NVIDIA.
• 1:05
  SPEAKER_00: It is a bet that enterprise customers will want choice, especially for inference, where cost per answer increasingly decides who wins.
• 1:14
  SPEAKER_05: Next, a memory maker's record quarter.
• 1:17
  SPEAKER_02: SK Heinex has posted the kind of quarter that explains why memory is now the tightest link in the AI supply chain.
• 1:25
  SPEAKER_02: The South Korean chipmaker reported record third quarter revenue of about 104.6 trillion one, roughly$79 billion, up more than 300% from a year earlier.
• 1:39
  SPEAKER_02: Operating profits soared to about 83.3 trillion one, an operating margin of 80%.
• 1:47
  SPEAKER_02: The driver is high bandwidth memory, the stacked chips that sit beside AI accelerators.
• 1:54
  SPEAKER_02: The company says its newest generation, HBM4, will overtake the previous generation in both shipments and sales this quarter.
• 2:03
  SPEAKER_02: SK Heinrich also announced a shareholder return program worth more than 100 trillion one, about$75 billion.
• 2:14
  SPEAKER_04: Now from Google Deep Mind.
• 2:16
  SPEAKER_04: Google Deep Mind has unveiled Genie 4, the next version of its world model.
• 2:21
  SPEAKER_04: Unlike a video generator, a world model simulates an environment frame by frame, so a user can walk through it and change it as it runs.
• 2:29
  SPEAKER_04: Genie4 builds those worlds from text or image prompts in real time, and for the first time, it generates native audio alongside the visuals.
• 2:38
  SPEAKER_04: Users can change the weather, move the camera, or add characters and objects mid-simulation, and characters can be driven by Gemini models.
• 2:46
  SPEAKER_04: Reports say scenes now hold together for around two minutes, roughly double the previous version.
• 2:52
  SPEAKER_04: Google is calling it an early research preview, available only to trusted testers, and says objects can still drift over longer sessions.
• 3:00
  SPEAKER_04: Turning to Washington.
• 3:02
  SPEAKER_01: Anthropic is cutting the price of Clawed for Government.
• 3:05
  SPEAKER_01: The General Services Administration has expanded its OneGov agreement with the company, giving public agencies a 40% discount on Clawed Enterprise, Clawed for Government, and the Clawed Developer Platform.
• 3:17
  SPEAKER_01: Pricing is based on usage rather than seats, and there is no minimum purchase.
• 3:22
  SPEAKER_01: For the first time, the deal reaches beyond federal agencies to state, local, tribal, and territorial governments.
• 3:30
  SPEAKER_01: Anthropic estimates the arrangement could save government buyers about$1 billion, and says roughly 500 government organizations already use Claude.
• 3:39
  SPEAKER_01: Direct purchases are available now, with purchases through cloud providers expected in the coming months.
• 3:45
  SPEAKER_01: The move comes while Anthropic's separate dispute with the Pentagon remains unresolved.
• 3:50
  SPEAKER_05: Next, a memo from Amazon.
• 3:52
  SPEAKER_03: Amazon is telling its engineers that more AI use is not automatically better.
• 3:58
  SPEAKER_03: In a memo published Tuesday, the company acknowledged that it has sometimes tracked token usage, AI tool adoption, and the volume of code changes because long-term outcomes are slow to show up.
• 4:10
  SPEAKER_03: But leaders warned that treating those numbers as targets can drive up cost and complexity.
• 4:16
  SPEAKER_03: The memo says, chasing token counts runs against Amazon's frugality principle, and counting lines of code runs against its push to invent and simplify.
• 4:26
  SPEAKER_03: Teams are told to measure outcomes instead.
• 4:29
  SPEAKER_03: It also says that if a person's work does not benefit from current AI tools, that is a finding, not a failing.
• 4:36
  SPEAKER_03: It is a notable shift in tone from one of the heaviest AI spenders in the industry.
• 4:41
  SPEAKER_05: Now to insurance technology.
• 4:44
  SPEAKER_00: Insurance software maker Guidewire is buying its way deeper into AI underwriting.
• 4:48
  SPEAKER_00: The company has signed a definitive agreement to acquire Hyper Exponential, a London-based platform that helps commercial property and casualty insurers price risk.
• 4:58
  SPEAKER_00: Guidewire will pay a base price of$1.7 billion in cash, plus up to$400 million more if recurring revenue targets are met.
• 5:07
  SPEAKER_00: Hyper Exponential describes itself as an agentic platform where AI agents work alongside human underwriters to assess, price, and decide on policies.
• 5:17
  SPEAKER_00: Its software already supports decisions covering more than$75 billion in annual commercial premium.
• 5:24
  SPEAKER_00: GuideWire announced the deal at its Connections conference in Las Vegas, and investors reacted cautiously, sending its shares down on the news.
• 5:32
  SPEAKER_05: Next, powering data centers.
• 5:34
  SPEAKER_02: Power, not chips, is increasingly the bottleneck for AI data centers, and Vertiv is spending to address it.
• 5:42
  SPEAKER_02: The infrastructure company has completed its acquisition of Utility Innovation Group for about$1.45 billion in upfront cash.
• 5:53
  SPEAKER_02: The deal includes up to$1.15 billion more in contingent payments tied to earnings targets over the next two years.
• 6:04
  SPEAKER_02: Utility Innovation Group specializes in microgrids, on-site power generation, and energy management.
• 6:11
  SPEAKER_02: Vertiv says the purchase extends its portfolio from the grid connection all the way to the chip without depending on any single generation technology or supplier.
• 6:22
  SPEAKER_02: The pitch is aimed squarely at operators who cannot wait years for utility hookups and need to bring their own power.
• 6:31
  SPEAKER_04: Turning to life sciences, Desolde Systemese is pulling its life sciences business onto one AI platform.
• 6:39
  SPEAKER_04: The French software company said it has completed its acquisition of Eris Global, a maker of drug safety and regulatory software, for about$1.8 billion in cash, with up to$200 million more tied to AI-related revenue goals.
• 6:54
  SPEAKER_04: Eris Global now joins DeSolt's BioVia Research Tools and its metadata clinical trial business.
• 7:00
  SPEAKER_04: Together, they are being united inside the company's 3D Universes platform, which DesO describes as an AI native system spanning the full drug life cycle, from molecule discovery to patient safety monitoring.
• 7:14
  SPEAKER_04: Dassault also named Dr.
• 7:16
  SPEAKER_04: Ashwini Zanuz to lead the combined life sciences and healthcare unit and raised its full-year objectives alongside third-quarter results.
• 7:24
  SPEAKER_04: Next, AI at the edge.
• 7:26
  SPEAKER_01: Akamai is turning its global edge network into a place to run AI.
• 7:31
  SPEAKER_01: The company has introduced two new products for Akamai Cloud, both in beta.
• 7:36
  SPEAKER_01: The first, Akamai Inference, serves open weight and custom models through interfaces compatible with open AIs, with either pay-per-use serverless options or dedicated deployments.
• 7:49
  SPEAKER_01: The second, Akamai Agent Sandbox, gives AI agents somewhere safe to execute the code they write.
• 7:56
  SPEAKER_01: Agents run on Akamai functions, and their generated code runs in a separate, isolated environment using WebAssembly with limits on compute, memory, and outbound connections.
• 8:07
  SPEAKER_01: The sandbox is being released as an open source part of the spin framework.
• 8:12
  SPEAKER_01: Akamai is betting that companies will want inference running close to their users and agents kept on a short leash.
• 8:18
  SPEAKER_05: Now, following the money.
• 8:20
  SPEAKER_03: The financing machinery behind AI compute keeps getting bigger.
• 8:24
  SPEAKER_03: Reuters reports that Argentum AI, a cloud computing firm backed by Supermicro, is in talks to secure about$5 billion in financing for AI chips from a major private credit firm.
• 8:36
  SPEAKER_03: According to people familiar with the matter, the money would pay for additional chips at a data center Argentum secured in Poland with DL Invest Group.
• 8:45
  SPEAKER_03: The debt is aimed at hardware leased to customers that do not have investment grade credit ratings.
• 8:51
  SPEAKER_03: Argentum has reportedly secured more than$12 billion in contracted revenue.
• 8:57
  SPEAKER_03: Talks are ongoing and terms could change.
• 9:00
  SPEAKER_03: Supermicro stock rose more than 7% on the report as investors weighed what more neo-cloud demand could mean for server makers.
• 9:08
  SPEAKER_05: Next, guarding AI agents.
• 9:11
  SPEAKER_00: As companies hand more work to AI agents, securing what those agents can touch has become its own market.
• 9:18
  SPEAKER_00: In Physical, a San Francisco security startup, has raised a$40 million Series B at a$600 million valuation.
• 9:26
  SPEAKER_00: The round was led by investor Alad Gill, who also backed the company's Series A with Gradient and Y Combinator participating.
• 9:34
  SPEAKER_00: In Physical Manages Secrets, the passwords, keys, and certificates that software uses to reach sensitive systems.
• 9:41
  SPEAKER_00: The company says it will put more investment into Agent Vault, a product that lets AI agents use credentials without ever seeing the underlying secrets.
• 9:51
  SPEAKER_00: The new funding brings in Physical's total raise to nearly$60 million and will expand its engineering and sales teams.
• 9:58
  SPEAKER_02: Finally, a bet on efficiency.
• 10:00
  SPEAKER_02: Nvidia is backing software that helps companies need fewer of its chips.
• 10:06
  SPEAKER_02: SimpliSmart, a Bengaluru-based AI inference startup, has raised$14 million in a Series B round led by Dallas Venture Capital, with NVIDIA, Accel, Shastra VC, and MicroMax participating.
• 10:23
  SPEAKER_02: The round reportedly values the four-year-old company at about$100 million.
• 10:29
  SPEAKER_02: Simply Smart Software helps enterprises deploy and run AI models efficiently on graphics processors and lets cloud providers turn idle GPU fleets into services they can sell.
• 10:44
  SPEAKER_02: The company says its annualized revenue run rate grew 14 fold over the past six months, though that reflects recent usage rather than contracted revenue.
• 10:57
  SPEAKER_02: More than half of its revenue comes from the United States.