Tech Brew Ride Home – August 22, 2025
Host: Brian McCullough
Main Theme:
The episode unpacks Nvidia's escalating challenges in maintaining access to the Chinese market after regulators push against its H20 AI chip, alongside major cloud partnerships, speculation around an OpenAI buyout, the real story behind ChatGPT's web search upgrades, and a look at the rise of AI-generated likenesses and the healthcare software giant Epic Systems.
Nvidia’s China Setback: H20 Chip Production Halted
[01:31 – 09:26]
- Production Stoppage:
Nvidia has instructed suppliers—including Amkor and Samsung—to suspend H20 chip production. This follows guidance from Chinese regulators for local tech firms to avoid the H20 due to alleged "security concerns" about potential chip backdoors and tracking. - Background & Market Implications:
- China is anxious about U.S. chips funneling data abroad and has ramped pressure on firms to buy homegrown AI chips (e.g., Huawei).
- Despite the Trump administration lifting a prior U.S. ban, China’s stance has left Nvidia’s China market goals “in limbo.”
- “Chinese authorities are encouraging local companies to use Chinese made AI chips such as those sold by Huawei.” – Brian [03:13]
- Nvidia’s Response:
- Official statement: “We constantly manage our supply chain to address market conditions... Allowing US Chips for beneficial commercial business use is good for everyone.” [03:46]
- Denial of backdoors: “The market can use the H20 with confidence.”
- Jensen Huang Addresses Fears:
- Nvidia CEO Jensen Huang confirmed talks with the Trump administration about a potential new 'B30A' chip for the Chinese market—a follow-on to the H20, but not Nvidia’s decision alone.
- “I’m offering a new product to China for AI data centers, the follow on to H20... but it’s too soon to know.” – Jensen Huang [04:18]
- The B30A would reportedly operate at about half the speed of their main B300 chips.
- On the uproar, Huang added: “There’s no such thing and there never has been [a backdoor]... Shipping H20 to China is not a national security concern. It’s great for America and it is great for the Chinese market.” [05:37]
- Chinese Progress on Domestic AI Chips:
- AI startup Deepseek claims new advances in local “AI stacking” chips could soon replace Nvidia products, fanning rumors of an AI hardware breakthrough.
Meta and Google Close $10 Billion Cloud Deal
[09:26 – 13:35]
- Key Details:
- Meta and Google have signed a $10B+, six-year cloud services agreement—one of Google Cloud’s largest-ever, providing Meta with storage, networking, and access to Nvidia GPUs in Google data centers.
- Meta still heavily invests in its own in-house data centers but diversifies with Amazon AWS, Microsoft Azure, Oracle, CoreWeave, and now Google Cloud.
- Strategic Implications:
- “Landing Meta as a major customer is arguably an even bigger feather in Google’s cap. Not only is Meta its biggest digital advertising rival, the companies compete on multiple other fronts...” [10:40]
- The deal fuels Google Cloud’s recent business momentum, attracting even competitors amid an AI data center arms race.
- Multi-cloud approach increases Meta’s leverage for future contract negotiations and delivers better global user service resilience.
Elon Musk–Zuckerberg & the OpenAI Takeover Bid
[13:35 – 15:18]
- Court Filing Reveals Approach:
- Elon Musk claims to have spoken to Mark Zuckerberg in February about backing a $97.4B OpenAI buyout, but neither party signed a letter of intent.
- “Any potential partnership between Musk and Zuckerberg would stand in stark contrast to previous interactions in 2023, when the pair taunted each other online...” [13:56]
- OpenAI requests Meta hand over documents related to the talks for Musk’s ongoing lawsuit; Meta refused, noting it never joined the bid.
OpenAI’s Use of Google Search via SerpAPI
[15:18 – 19:47]
- How ChatGPT Handles Current Events:
- OpenAI sources current-events data for ChatGPT responses via SerpAPI—a web-scraping service pulling Google search results.
- Google previously denied OpenAI direct API access but has not blocked SerpAPI legally, possibly due to ongoing antitrust litigation and regulatory scrutiny.
- “Google executives have privately derided SERP API... tried various techniques to make it harder for the firm to scrape high quality information...” [17:15]
- OpenAI’s Data Mix & Industry Web:
- OpenAI doesn’t rely exclusively on Google; it also uses Microsoft Bing API, its own web crawler, and similar services from Brave and Exa.
- Apple, Perplexity AI, and others are also listed as SerpAPI customers.
- OpenAI: “We retrieve accurate, contextually relevant information from web pages and a variety of providers. This allows us to surface and synthesize information from multiple sources.” – OpenAI spokesperson [18:39]
- Replicating Google’s depth for rare queries is considered “extremely difficult” for others, according to OpenAI execs.
Weekend Long Read Highlights
[19:47 – 25:55]
1. Selling Your Digital Likeness to AI
- Story:
Actor Scott Jackman sold his likeness to TikTok for a one-off payment; now his digital avatar appears across the internet, often in ways he never approved (e.g., ads in languages he doesn’t speak, without his signature facial hair). - Performer Perspective:
- “You really don’t know the ramifications of this.” — Scott Jackman
- Artist Tracy Federer: “I might as well get paid for it before they decide not to pay anyone.”
- Broader Implications:
- Marketers can mass-test scripts across digital personas, massively driving down costs.
- NYU law professor Gene Fromer warns contracts lag technology, leaving talent open to abuse.
- Venture capitalist Joe Marchese predicts brands will prize cost savings but risk consumer backlash as AI deepfakes become widespread.
2. Epic Systems Profile
- Epic’s Origins & Scale:
- Founded in 1979 by Judy Faulkner with $70k; now controls 42% of U.S. acute care hospital software and reaches 325 million patients.
- Faulkner, now 82, insists Epic remain private to prioritize “purpose over profits.”
- Corporate Culture & Controversies:
- Epic’s campus features whimsical designs (Oz, Alice in Wonderland).
- Company culture prioritizes integrity and customer-centricity; Faulkner is known for swift, detail-oriented leadership.
- Critics cite clunky user interfaces and anti-competitive tactics, but strong mission focus remains.
- Succession plan involves a trust to safeguard Epic’s independence, along with a major philanthropic commitment through the Roots and Wings Foundation.
Notable Quotes & Moments
- Jensen Huang on China tensions:
“There’s no such thing and there never has been [a backdoor]... Shipping H20 to China is not a national security concern.” [05:37] - On Multi-Cloud Era:
“Landing Meta as a major customer is arguably an even bigger feather in Google’s cap...” [10:40] - On digital likeness risks:
“You really don’t know the ramifications of this.” — Scott Jackman [20:27] “I might as well get paid for it before they decide not to pay anyone.” — Tracy Federer [21:02] - Faulkner’s Mission:
“Purpose over profits” attitude at Epic [23:03]
Key Timestamps
- 01:31 – Nvidia’s H20 chip production halt and China policy details
- 04:18 – Jensen Huang on B30A chip and U.S. approval
- 09:26 – $10B Meta–Google Cloud deal
- 13:35 – Musk–Zuckerberg OpenAI takeover discussions revealed
- 15:18 – OpenAI’s use of SerpAPI and industry implications
- 19:47 – Weekend Long Reads: AI avatars, Epic Systems profile
Summary Takeaways
- U.S.-China Tech Friction: Nvidia’s China saga highlights the deepening rift—and the country’s fast progress on homegrown alternatives.
- Big Cloud Alliances: Despite fierce rivalry, even Meta turns to Google Cloud for world-scale AI ambitions, signaling new dynamics in data infrastructure.
- AI-generated Likenesses: Early digital actors are both reaping and regretting the bargain, as their faces are used far and wide—with little recourse.
- Behind ChatGPT’s News Savvy: Current events answers may come courtesy of search-scraping, not a direct Google API.
- Healthcare Tech Titan: Epic Systems’s founder keeps faith with vision and philanthropy, as the company quietly shapes care for millions.
