Latest / DX Today | No-Hype Podcast & News About AI & DX / AI News Today: Marvell's $115B Lumentum Megadeal

Transcript

AI News Today: Marvell's $115B Lumentum Megadeal · DX Today | No-Hype Podcast & News About AI & DX

• 0:00
  SPEAKER_03: It's Tuesday, November 3, 2026.
• 0:03
  SPEAKER_03: You're listening to the DX Today AI Daily Brief.
• 0:07
  SPEAKER_03: Today, Marvel agrees to buy Lumentum in a$115 billion bet on optical networking.
• 0:14
  SPEAKER_03: Anthropic's amended IPO filing shows nearly$17 billion in quarterly revenue alongside almost$900 billion in commitments.
• 0:25
  SPEAKER_03: And a top advisor to Europe's highest court says Google's Gemini may infringe the rights of news publishers.
• 0:32
  SPEAKER_03: Let's get into it.
• 0:33
  SPEAKER_00: Marvel Technology announced on Monday that it will acquire Lumentum Holdings in an all-stock deal valued at about$115 billion, including debt.
• 0:42
  SPEAKER_00: Lumentum makes the lasers and optical components that move huge volumes of data between servers, and Marvell is betting that light, not copper, will carry the next wave of AI data center traffic.
• 0:54
  SPEAKER_00: The New York Times reports it would be the largest deal ever by value between two publicly held technology companies.
• 1:00
  SPEAKER_00: The offer is a premium of about 24% to Lementum's price before talks leaked last week.
• 1:06
  SPEAKER_00: When the deal closes, Marvell's shareholders will own roughly two-thirds of the combined company.
• 1:12
  SPEAKER_00: Reuters notes the merger is designed to help Marvell compete more aggressively with Broadcom and custom chips and networking.
• 1:18
  SPEAKER_03: Next, a blockbuster IPO filing.
• 1:21
  SPEAKER_04: Anthropic has filed an amended IPO registration with the Securities and Exchange Commission adding its third quarter numbers.
• 1:31
  SPEAKER_04: Quarterly revenue reached a little under 17 billion against roughly 1.4 billion in the same quarter last year.
• 1:41
  SPEAKER_04: But growth from the prior quarter slowed to about 47%, down from 144% in the second quarter.
• 1:50
  SPEAKER_04: The company reported a net loss of roughly 23 billion and an operating loss of 567 million after posting an operating profit the quarter before.
• 2:02
  SPEAKER_04: The filing also lists about$895 billion in purchase and lease commitments, mostly for compute and infrastructure.
• 2:12
  SPEAKER_04: The largest is a five-year equipment lease with Broadcom, valued at about$161 billion.
• 2:19
  SPEAKER_04: Reuters reports the filing flags new risks around securing that capacity.
• 2:36
  SPEAKER_02: Earnings per share rose 14% from a year earlier.
• 2:45
  SPEAKER_02: And services hit a record of about$31.5 billion.
• 2:50
  SPEAKER_02: The outlook was more cautious.
• 2:52
  SPEAKER_02: Apple expects revenue growth of 7-10% in the holiday quarter, with gross margin of 45 to 46%.
• 2:59
  SPEAKER_02: The company pointed to soaring memory prices and the costs of ramping its new AI-powered Siri as pressures on margins.
• 3:07
  SPEAKER_02: Shares fell more than 4% in pre-market trading on Tuesday.
• 3:11
  SPEAKER_02: Another strong quarter.
• 3:12
  SPEAKER_01: Palantir delivered its first quarter with more than$2 billion in revenue.
• 3:17
  SPEAKER_01: Third quarter sales reached about$2.31 billion, up 96% from a year ago, and well above analyst estimates.
• 3:25
  SPEAKER_01: The milestone that caught investors' attention is a shift in the mix of the business.
• 3:30
  SPEAKER_01: For the first time, revenue from American commercial customers at$984 million topped revenue from the United States government at$953 million.
• 3:42
  SPEAKER_01: American commercial revenue grew 148% as more companies adopt the company's AI platform.
• 3:50
  SPEAKER_01: Total revenue in the United States rose 119%.
• 3:54
  SPEAKER_01: Palantir raised its full-year revenue guidance to about$8.53 billion, and its fourth quarter outlook came in above consensus.
• 4:03
  SPEAKER_03: Now AI in the courtroom.
• 4:22
  SPEAKER_05: He said a chatbot's reuse of press articles can violate publishers' rights when it reproduces recognizable expression from the original work or when the content was accessed unlawfully.
• 4:34
  SPEAKER_05: Simply reporting the facts contained in an article would not qualify.
• 4:39
  SPEAKER_05: The opinion is not binding, and the court's judges will issue their own ruling later.
• 4:44
  SPEAKER_05: But it is the first time a case at the European court has squarely addressed AI outputs and publisher rights, and the judges often follow their advisors' lead.
• 4:54
  SPEAKER_03: Over to the Gulf.
• 4:55
  SPEAKER_00: OpenAI has signed a multi-year partnership with the Abu Dhabi National Oil Company, known as ADNOC, and the technology group G42.
• 5:04
  SPEAKER_00: ADNOC says it will become the first energy company in the world to deploy OpenAI Frontier, a platform for connecting and coordinating AI agents across complex business operations.
• 5:15
  SPEAKER_00: The system will run on OpenAI's models hosted on G42's sovereign infrastructure inside the United Arab Emirates.
• 5:23
  SPEAKER_00: Financial terms were not disclosed.
• 5:25
  SPEAKER_00: The deal was announced at the Adipec Energy Conference, where Chief Executive Sam Altman appeared.
• 5:31
  SPEAKER_00: Abu Dhabi's media office says four agreements with OpenAI were signed on Monday, covering oil field planning, government compliance, and investment analysis.
• 5:40
  SPEAKER_00: Altman described the Emirates as the most forward-looking country in the world on AI adoption.
• 5:45
  SPEAKER_04: Next, a voice AI deal.
• 5:47
  SPEAKER_04: Salesforce has signed a definitive agreement to acquire Cresta, a company that builds voice agents for customer service.
• 5:55
  SPEAKER_04: Cresta's software can resolve complex customer calls from start to finish, and it also guides human representatives in real time during conversations.
• 6:06
  SPEAKER_04: Salesforce plans to fold the technology into Service Cloud and its Agent Force platform, extending its AI agents from chat and email into the phone channel.
• 6:19
  SPEAKER_04: Salesforce did not disclose the price.
• 6:22
  SPEAKER_04: The information, citing a person familiar with the deal, reported that the agreement values Cresta at about$2.7 billion.
• 6:32
  SPEAKER_04: Voice remains one of the largest and most expensive parts of the contact center business and a growing battleground for enterprise software companies.
• 6:42
  SPEAKER_02: Now to China.
• 6:44
  SPEAKER_02: Chinese delivery giant Maitwan has released Longcat 2.5, a native multimodal model built for long-running agent tasks.
• 6:53
  SPEAKER_02: The model has 1.6 trillion total parameters, with about 48 billion active for each token it processes.
• 7:00
  SPEAKER_02: Maituan says the model was trained entirely on domestic Chinese computing hardware, extending its effort to build frontier scale AI without relying on American chips.
• 7:11
  SPEAKER_02: The release adds visual understanding to the Longcat family, allowing agents to work with images as well as text.
• 7:18
  SPEAKER_02: Maituan's move comes as Chinese technology companies race to prove that homegrown accelerators can handle training at this scale.
• 7:26
  SPEAKER_02: At a time when access to the most advanced American processors remains restricted by export controls.
• 7:33
  SPEAKER_02: Next, a memory milestone.
• 7:35
  SPEAKER_01: Qualcomm says its new memory technology for AI inference has passed its first hardware tests.
• 7:41
  SPEAKER_01: The company validated first-generation high bandwidth compute silicon in its San Diego lab, running the full memory read and write path along with computations performed next to the memory itself.
• 7:52
  SPEAKER_01: In testing, the design delivered about 133 terabytes per second of memory bandwidth on its Dragonfly AI 250 accelerator card, roughly 18 times what its current card achieves with conventional mobile memory.
• 8:06
  SPEAKER_01: Qualcomm argues that inference is increasingly limited by how fast data moves, not by raw compute.
• 8:13
  SPEAKER_01: The Dragonfly AI 250 is expected to begin sampling to customers in the first half of 2027, part of Qualcomm's push into the data center.
• 8:22
  SPEAKER_03: More on power for AI.
• 8:24
  SPEAKER_05: Chipmaker on SEMI has agreed to acquire Farric, a developer of integrated voltage regulators, for$600 million in cash, plus up to$175 million more tied to revenue milestones.
• 8:37
  SPEAKER_05: Ferric's technology places power regulation very close to the processor itself, which helps deliver clean, efficient power to the most demanding AI chips.
• 8:47
  SPEAKER_05: As accelerators draw more and more electricity power delivery has become one of the hardest engineering problems inside the data center.
• 8:55
  SPEAKER_05: On SEMI says the purchase strengthens its position in power for next generation AI processors.
• 9:02
  SPEAKER_05: The acquisition will be funded with cash on hand and is expected to close by the end of the year, subject to customary conditions.
• 9:09
  SPEAKER_03: Now, AI and jobs.
• 9:20
  SPEAKER_00: In an internal memo first reported by the register, Chief Executive Dirk Peter Van Leeuwen told employees that automation and AI have reduced the need for some existing enrolls.
• 9:31
  SPEAKER_00: The memo did not say which teams will be affected.
• 9:34
  SPEAKER_00: SUS is best known for its enterprise Linux distribution and its container management tools.
• 9:39
  SPEAKER_00: And like many software companies, it is reshaping itself around AI features.
• 9:44
  SPEAKER_00: The announcement adds to a growing list of technology employers tying job cuts explicitly to automation rather than to broader economic conditions as companies rethink how many people they need to build and support their products.
• 9:57
  SPEAKER_04: And finally, defense AI.
• 9:59
  SPEAKER_04: Cohere has launched a Defense AI accelerator in Singapore.
• 10:03
  SPEAKER_04: The Canadian company, which focuses on secure AI for governments and regulated industries, says the new center is designed to help defense organizations and other sovereign customers move from experimenting with AI to deploying it in operations.
• 10:21
  SPEAKER_04: The work is focused on highly secure environments, where systems often need to run inside an organization's own infrastructure rather than on a public cloud.
• 10:33
  SPEAKER_04: Singapore gives Cohere a regional base close to defense and government buyers across Asia.
• 10:41
  SPEAKER_04: The move reflects a wider trend as AI developers compete for public sector contracts by offering models that governments can control, audit, and keep within their own borders.