OpenAI, Anthropic, and Google DeepMind just confirmed they're coordinating on AI safety, then disagreed on whether they need an antitrust waiver to do it. @PatrickMoorhead calls the irony out directly. @danielnewmanUV draws the sharper line: public companies are moving cautiously; VC-backed labs are full steam ahead.

A call to slow the frontier landed in a market that keeps buying compute. Frontier labs want regulatory cover; open models, custom enterprise models, and compute pricing power are quietly redrawing where the AI moat sits. @PatrickMoorhead and @danielnewmanUV break down the pacing debate, Salesforce Dreamforce updates, OpenRouter share shifts, and OpenAI's $1.2T question on Ep. 320 of The Six Five Pod. Tune in: 00:00:00 - Introduction and Milestone Celebration 00:06:20 - ‘We Must Pace the Frontier’ Essay Discussion 00:15:45 - OpenAI, Anthropic, and Google Collaboration 00:23:44 - Salesforce's Dreamforce Insights 00:33:15 - OpenAI vs Anthropic’s Market Positions and API Traffic 00:40:55 - AI Infra Summit Highlights 00:44:08 - The Flip: OpenAI's $1.2T Valuation 00:52:45 - Doomer Sell-off Flips Into a Chip Rally. 00:55:40 - Salesforce New FY30 Revenue Target 00:59:40 - Nebius’ Pricing Moves

The AI agents most likely to expose sensitive data may be the ones no one knows exist. At The Six Five Summit, Veeam CEO @anandeswaran explained why @Veeam makes data the control point, not the agent. If an agent is overprivileged, unknown, or rogue, protected data never enters its context window. Watch:

AI frameworks prioritize FLOPs and TOPS. But @SamsungSemiUS President Paul Cho says the bottleneck has shifted. Memory bandwidth now determines AI inference performance and cost. He believes memory-centric compute could define AI’s next decade. Watch:

Broadcom's Michael Jordan says “Harvest Now, Decrypt Later” isn’t a future threat. Governments and organizations with the resources to do so are already stockpiling encrypted data for the quantum era. @MattKimball_MIS talks with Jordan, who is a Distinguished Engineer for Cybersecurity @Broadcom, about what a real post-quantum migration timeline looks like, and why cryptographic discovery must begin before the deadline is clear. $AVGO

Idle GPUs waiting on hard drives are burning capital, says @wd_corporation SVP Tim Rausch. At AI Infra Summit 2026, @MattKimball_MIS spoke w/ Rausch about why storage may determine how far AI can scale. They are targeting: -60-terabyte drives by 2028 -100-terabyte drives by 2030 -Read/write technology delivering up to 8x today’s throughput $WDC

Tungsten resistance spikes at 2 nanometers. @Applied4Tech built Spectral Moly ALD to replace it with molybdenum, cutting resistance roughly 15%. Mukund Srinivasan told @PatrickMoorhead at the Six Five Summit why the next gains in AI efficiency start at the materials layer. $AMAT

The AI infrastructure opportunity doesn’t stop at the XPU. @MarvellTech’s SVP Will Chu says “the networking, memory, storage, and security around a chip means 3 - 4 more “XPU attach” design wins per custom XPU. Dave Lazovsky (fmr. CEO of and now SVP, says reasoning models are why: KV cache is up 10x in 9 months, pushing scale-up past a single rack. Cumulative data center capex forecasts through 2030 climbed from $10.7 trillion to $12 trillion within a single quarter. Chu and Lazovsky’s combined perspectives make up the system Marvell is counting on to capture that spend. $MRVL

Enterprise AI has to drive growth without letting costs spiral. At The Six Five Summit, @Snowflake CEO @RamaswmySridhar shared that their AI sales agent serves all 4,000 salespeople and costs less than the dashboarding tools it replaced. Watch the full session:

68% of leaders worry their AI efforts will fail because they aren’t embedded in their organization's workflows. In Ep. 6 of The Orchestrators with @IBM, Salima Lin of @IBMConsulting explains why companies must rethink roles, reskill people, and redesign how humans and AI work together, or risk being stuck in endless pilots. Watch:

Google Cloud now processes more than three quadrillion tokens a month, 7x more than a year ago. Mark Lohmeyer, VP & GM of AI and Computing Infra at @googlecloud, tells @danielnewmanUV why only 17% of enterprises believe their infrastructure can handle the agentic AI load they expect. $GOOGL

AI can now find decades-old mainframe vulnerabilities in seconds, and defenders don't get the same margin for error attackers do. Closing that gap requires faster SECINT cycles and experienced human judgment, not just faster discovery. @danielnewmanUV and @GregLotko sit down with @Broadcom’s Chad Rikansrud to break down what frontier AI means for mainframe security. $AVGO

Jim Anderson, CEO of @CoherentCorp, gave his leadership team a two-word mandate this year: go faster. He tells @PatrickMoorhead during The Six Five Summit that Coherent is quadrupling its Indium Phosphide output capacity over the next 24 months to keep pace with data center demand, and why NVIDIA CEO Jensen Huang joined @CoherentCorp’s Texas groundbreaking. $COHR

Behind the scenes at AI Infra Summit! @MattKimball_MIS and @BrendanBurkeX sat down with leaders from @Ambarella_Inc, @wd_corporation, @Synopsys, and @MiTACcomputing to explore edge AI, storage, AI engineering, and scalable systems. Full interviews coming soon!

Broadcom pointed frontier AI models at its own open-source code and saw vulnerability reports jump 2,000% before its own scan even started. At @VMwareExplore, Purnima Padmanabhan tells @PatrickMoorhead and @danielnewmanUV why that's driving @Broadcom's new TrueSource security brand. $AVGO

A federal advisory called out six Chinese AI labs for conducting aggressive, industrial-scale knowledge distillation but banned nothing. Is that enough to chill enterprise adoption? @danielnewmanUV argues NSA, CISA, and FBI scrutiny will make legal teams think twice. @PatrickMoorhead counters that rejecting open-weight models with up to 90% cost savings over vague guidance is an overcorrection. Note: this is Six Five's simulated "Flip" debate segment where the positions argued are assigned randomly and do not reflect hosts' personal views.

Physical AI prototypes are everywhere on trade show floors. Ambarella measures readiness by a different standard: fleets running reliably for years, upgraded and monitored across thousands of deployed units. At AI Infra 2026, @BrendanBurkeX and @MattKimball_MIS talk with Muneyb Minhazuddin, Customer Growth Officer at @Ambarella_Inc, about a 22-year chip architecture cutting memory use 10x. $AMBA

3M's sales reps are saving 5+ hours per person after adopting Amazon Quick. A corporate legal team is saving millions by replacing quarterly patent and regulatory scans with a long-running agent. @PatrickMoorhead speaks with Jigar Thakkar, VP of Agentic AI for Business at @awscloud, to break down why culture and data connectivity separate enterprises scaling agentic AI from those stuck in pilots.

79% of executives expect AI to drive significant revenue by 2030. Only 24% know what those revenue streams will be. Salima Lin calls that gap the problem CEOs need to solve first. @danielnewmanUV talks with Lin and Dalia Tarabay of @IBM about why 68% of leaders worry their AI efforts will fail from lack of integration, and why almost two-thirds of CEOs say they'll take on more risk than their competitors to keep an edge.

A modern smartphone chip has tens of billions of transistors, 60 miles of wiring, is produced by more than a thousand manufacturing steps, and costs about $30. @Applied4Tech CEO Gary Dickerson tells @PatrickMoorhead and @danielnewmanUV that's the kind of miracle AI compute is about to demand more of, faster than ever before. $AMAT