Tech Brew Ride Home - Blackouts Take Waymo Out

Episode Date: December 22, 2025

Turns out when the lights go out, Waymo’s don’t handle that well. Larry Ellison actually puts his money on the line. Somebody is pirating music like it’s 1999. And two deep-dive looks at whether or not Google’s TPU’s really are a threat to Nvidia and OpenAI. Waymo resumes robotaxi service in San Francisco after blackout chaos — Musk says Tesla car service unaffected (CNBC) Paramount guarantees Larry Ellison backing in amended WBD bid (CNBC) Instacart Scraps All Price Tests After Customer Pushback (WSJ) Spotify Music Library Scraped by Pirate Activist Group (Billboard) ChatGPT will now let you pick how nice it is (The Verge) TPU Mania (The Chip Letter) Why Nvidia maintains its moat and Gemini won’t kill OpenAI (SiliconANGLE) Learn more about your ad choices. Visit megaphone.fm/adchoices

It's peak pollination season, and my business is scaling fast. To keep the nectar flowing, I need a phone plan with top priority data speed. That's why I chose GoogleFi Wireless. My connections stay strong even when the hive is buzzing. Plus, unlimited plans start at $35 a month. Now, that's a deal that doesn't stay. Explore Google Fi Wireless plans today. Plus taxes and government fees.

Google Fi Wireless is not subject to data traffic deprioritization during times of high network usage. Welcome to the TechBrew ride home for Monday, December 22nd, 2025. I'm Brian McCullough today. Turns out when the lights go out, Waymo's don't handle that well. Larry Ellison actually puts his money on the line. Somebody is pirating music like it's 1999, and two deep dive looks at to whether or not Google's TPUs really are a threat to Nvidia and OpenAI. Here's what you miss today in the world of tech. You may have noticed that your CRM system has become crazy with bots and agents. the TechBrew Ride Home, we love AI, but not when it's used to commit fraud or launch ransomware. Fortunately, Momoto, the creator of continuous verification, now offers the next generation

of Continuous Captcha. Momoto Continuous Captcha diverts only the agents and bots attempting to gain access to your CRM through registration forms. With Momoto, good agents can do what agents do best, enable your customers to have more productive experiences on your website or app, while stopping bad agents from defrauding you or your customers. It's a win for everyone except the cybercriminals. Right now, our listeners can purchase a year of Momoto Continuous Capture for $5,000, a 20% discount

on their lowest price plan. To learn more, head to mimoto.aI slash ride home. That's mimoto. com.com. Waymo has resumed its Roboto taxi service in San Francisco after suspending it during a recent power blackout. But they say most trips did finish before vehicles returned to depots or pulled over, quoting CNBC.

Yesterday's power outage was a widespread event that caused gridlock across San Francisco with non-functioning traffic signals and transit disruptions, a Waymo spokesperson, Suzanne Phileon told CNBC in an emailed statement Sunday afternoon, while the failure of the utility infrastructure was significant, we are committed to ensuring our technology adjusts to traffic flow during such events, she added. As power outages spread yesterday, videos shared on social media appeared to show multiple Waymo vehicles stalled in traffic in different parts of the. city. San Francisco resident Matt Schoolfield said he saw at least three Waymo autonomous vehicles

stopped in traffic Saturday around 9.45 p.m. local time, including one he photographed on Turk Boulevard near Parker Avenue. They were just stopping in the middle of the street, Schoolfield said. The power outages began around 109 p.m. Saturday and peaked roughly two hours later, affecting about 130,000 customers, according to Pacific Gas and Electric. As of Sunday morning, about 2,100 customers remained without power, mainly in the Presidio, the Richmond District, Golden State Park, and parts of downtown San Francisco. While the Waymo driver is designed to treat non-functional signals as four-way stops, the sheer scale of the outage led to instances where vehicles remained stationary longer than usual to confirm the state of the affected intersections.

Waymo's Philean also told CNBC, this contributed to traffic friction during the height of the congestion. Waymo, quote, closely coordinated with San Francisco City officials, she said, said and proactively paused its service as of Saturday evening and in the first half of the day on Sunday. The majority of active trips were successfully completed before vehicles were safely returned to depots or pulled over, she noted, end quote. Aramount says Larry Ellison, quote, agreed to provide an irrevocable personal guarantee of $40.4 billion for Paramount's WBD bid after WBD's board raised concerns about the offer. Yeah, I heard about this on the Matt Levine podcast. The whole Paramount bid is being backstopped by Larry Ellison's personal fortune, but weirdly, not directly, at least not initially. Like instead of saying, here's a check, I'll sign it, put it in escrow. Ellison said his blind trust would pay up if required. But why not just cut the check? Quoting CNBC. Paramount said Ellison, the father of Paramount CEO David Ellison, has also agreed not to revoke the Ellison family trust or adversely transgender.

transfer its assets during a pending transaction. Last week, Warner Brothers Discovery Chairman Samuel DiPiazza told CNBC's David Faber the board had concerns about the supposed backing of Oracle's co-founder Larry Ellison in the bid. We were not confident that one of the richest people in the world would be there at closing, D. Piazza said at the time, doing a deal is great. Closing a deal is better. WBD earlier this month agreed to sell its studio and streaming assets to Netflix in a transaction valued at roughly $83 billion on an enterprise basis. Paramount wants to buy the entirety of WBD, including its portfolio of TV networks, and says its offer comes with an enterprise value of $108.4 billion.

Paramount notably did not increase its bid on Monday, reiterating that it believes the deal is superior, though Paramount did increase its proposed reverse breakup fee to match that of Netflix's offer. What we've done in this amended filing is we've cleared the brush of obfuscation around the offer said Jerry Cardinali, founder and managing partner of Redbird Capital Partners on CNBC Squatbox on Monday. Cardinali said Monday that as part of the amended filing, Larry Ellison would back the bid through an irrevocable trust which is anchored by 1.2 billion Oracle shares. Like we've done through the six bids that we've made, we are being responsive to what their concerns

are, Cardinali said, end quote. Instacart says it is ending all item price tests effective immediately and won't charge different prices for the same items after a report into that practice that we spoke about last week. Quoting the journal, the change comes in response to customer feedback after a report earlier this month from Consumer Reports, Think Tank, Groundwork Collaborative and Media Outlet, More Perfect Union said the San Francisco grocery delivery company displayed several different prices for users who added the same item from the same store at the same time. The report said that the experiment using 437 shoppers across four cities who added the same items simultaneously to their

Instacart shopping carts from the same store, saw an average difference of 13% between the highest and lowest prices, with some differences as high as 23%. Instacart earlier this month said that the tests helped retailers understand consumer preferences. On Monday, a spokesperson said that the test fell short of expectations. At a time when families are working hard to stretch their grocery budgets. Customers should never have to question the prices they see on Instacart, the spokesperson said. Instacart has disputed characterizations that tests were a form of dynamic pricing, saying that the prices didn't change in real time or that the prices were a form of surveillance pricing that used personal demographic or behavioral data. Instead, Instacart said the tests were similar to how

retailers run price tests between different stores. The change also came after Instacart last week was ordered to pay $60 million in refunds to customers to settle allegations from the Federal Trade commissioned that it used deceptive practices to raise costs for shoppers. The FTC alleged that the grocery delivery platform falsely advertised free delivery and 100% satisfaction guarantees, and also didn't adequately disclose the terms for Instacart Plus membership, end quote. 1999 is calling, and it wants you to know that Torrance are back. Pirate activist group Anna's Archive says it scraped 86 million music files and 256 million rows of track metadata from Spotify and released all this in around 300 terabytes of torrent files.

Quoting Billboard. Reactions to the initial report by Anna's archive, like one circulating via a LinkedIn post from the CEO co-founder of ThirdShare, a startup that uses AI to build legal tools for media companies theorize, quote, anyone can now, in theory, create their own personal free version of Spotify, all music up to 2025, with enough storage and a personal media streaming server like Plex. The only real barriers are copyright law and fear of enforcement. Spotify's total audio files exceeded the number mentioned by Anna's archive. Still Zimmerman's commentary points out that the incident could potentially dwarf the largest previously available open music archive, music brains, which contains around 5 million unique tracks. Anna's archive, which

typically focuses on books and papers, said the project is part of its mission of, quote, preserving humanity's knowledge and culture and described the Spotify screen. as an effort to, quote, build a music archive primarily aimed at preservation. The post added, of course, Spotify doesn't have all the music in the world, but it's a great start. A Spotify spokesperson just released the following statement. Spotify has identified and disabled the nefarious user accounts that engaged in unlawful scraping. We've implemented new safeguards for these types of anti-copyright attacks and are actively monitoring for suspicious behavior. Since day one, we have stood with the artist community against piracy, and we are actively working

with our industry partners to protect creators and defend their rights, end quote. Quoting Rob Lathurn on X, the metadata is super interesting. Around 37% of songs represent around 99.6% of listens, so the remaining 63% of songs represent only around 0.4% of listens. They note Spotify doesn't publish stream counts for songs with less than 1,000 streams, forcing estimation slash error bars in the district. distribution part with around 70% of songs, end quote.

Study and play. Come together on a Windows 11 PC. And for a limited time, college students get the best of both worlds. Get the unreal college deal, everything you need to study and play with select Windows 11 PCs. Eligible students get a year of Microsoft 365 premium and a year of Xbox GamePass Ultimate with a custom color Xbox wireless controller. Learn more at Windows.com slash student offer. While supplies last, ends June 30th, terms at AKA.m.m.m.m.com. Ready to soundtrack your summer? With Red Bull Summer All Day Play, you choose a playlist that fits your summer vibe the best.

Are you a festival fanatic? A deep-end DJ, a road dog, or a trail mixer. Just add a song to your chosen playlist and put your summer on track. Red Bull Summer All Day Play. Red Bull gives you wings. Visit Red Bull.com slash Bright Summer ahead to learn more. See you this summer. Ambition comes in all shapes and sizes. At First Citizens Bank, we roll with your goals because we're built for what you're building. Fit for your ambition for Citizens Bank. OpenAI has rolled out an update to let users adjust ChatGPT's characteristics, including how warm or enthusiastic it is, also its use of headers, lists, and emojis, quoting TechCrunch.

OpenAI will now give you the ability to dial up or down ChatGPT's warmth and enthusiasm, An update rolling out on Friday gave you the option to choose whether you wanted more or less of these particular personality traits or stick with the default. There are also options to tweak how often chat GPT uses emoji, headers, and lists. You can find these settings by tapping the menu at the top left corner of the chat GPT app, selecting your profile, heading to personalization, and choosing ad characteristics. Alongside these options, you can pick a personality for the AI chatbot, including whether it's quirky, professional-friendly, cynical, and more. Another update changes the way you write emails with chat GPT as you can now update and format text directly inside the chat. You can also highlight chunks of text and ask chat GPT to make specific

changes to it rather than attempting to direct chat GPT to that section in a separate prompt, end quote. Finally today, two pieces that take a look at the hole are TPUs, a threat to NVIDIA's GPUs debate. A substack called the chip letter took a look at Google's TPU evolution. It's decision to sell chips to competitors after 12 years of internal use and the TPU versus GPU Architecture Showdown. In a piece titled, TPUMania, the author Babbage, says that at the heart of the excitement is Google's decision to use TPUs exclusively to train its latest Gemini 3 large language model, plus its choice to sell TPU hardware to external customers, reportedly including meta, a former rival, of course. These moves have triggered a noticeable vibe shift in the industry with

investors and commentators speculating about a structural change in the balance of power between GPU-centric and TPU-centric AI infrastructure. Alphabet stock, as we told you, has reflected that optimism. Babbage begins by grounding the TPU's rise in a long history, one that spans back over a decade. Google's TPU program actually began in 2013, long before anyone had heard of chat GPT or the massive boom in AI investment. The TPU concept itself draws on ideas from the late 70s about systolic arrays, networks of simple processors that rhythmically pump data through a computation pipeline ideal for the matrix multiplications at the core of deep learning. Although those ideas weren't widely used in earlier decades, by the 2010s the practical

need for a highly efficient neural network compute made them valuable again. The first TPU deployed in 2015 was relatively simple, a co-processor attached via PCIE that could only do integer inference tasks. but it was cheap, scalable, and highly efficient, and it became a workhorse within Google's data centers. Babbage humorously notes that TPUV1 launched a thousand chips, a reference to how its release sparked a proliferation of specialized AI accelerator efforts across the industry. Since then, successive generations of TPUs have evolved to support training and inference at scale, culminating in TPUV7, codenamed Ironwood, the latest flagship hardware. This new TPU generation sits at the center of the recent discussions with analysts, and researchers publishing detailed explorations of everything from hardware stack design and software

ecosystem issues to supply chain and interconnect innovations. Babbage curates some of the highlights emphasizing that the discussion around TPUV7 extends beyond raw performance to the broader ecosystem implications. One of the most important points is that the competition between TPUs and GPUs isn't just about which architecture is inherently better. Traditional narratives assume GPUs are best for training and TPUs are best for inference. but Google's use of TPUs for training leading models undermines that simplistic view. What will determine long-term market share is not just architectural merit, but execution quality, software support, ecosystem adoptions, and strategic positioning.

He also suggests that there's likely to be a convergence over time. GPUs have steadily added matrix acceleration features and dropped some higher precision support, bringing them closer to TPU style specialization. Rather than one clear winner, the future landscape may involve a mix of approaches each optimized for different segments of the AI workload spectrum. In this sense, the GPU and TPU debate resembles the classic CISC versus risk architectural debates of the past, not a single vector, but a dynamic tension shaping the direction of computer architecture for years to come. Then in the second piece I'm linking to Silicon Angle takes a deep dive look at this from a different angle,

saying Google's TPUs and Gemini are competitive, but NVIDIA's and OpenAI's first mover advantage makes their lead seem durable. Also, Google faces the dreaded innovator's dilemma. Silicon angle claims, at the heart of this debate, are two intertwined narratives circulating on Wall Street and in the Valley. Number one, Nvidia's architectural lead is slipping, that alternatives like Google's TPUs or other A6 will disrupt the GPU hegemony. And two, Google's Gemini is a fundamental threat to Open AIS. search and large language model leadership. Both assertions carry rhetorical weight, but neither holds up cleanly under scrutiny. Let's start with Nvidia's moat. Critics argue that because

alternatives like tensor processing units exist, and because models like Gemini have been trained on them, Nvidia's competitive advantage is crumbling, but this perspective misses a couple of critical realities. First, dominance in AI Silicon isn't just about raw micro-architectural performance. It's also about scalability, ecosystem, supply chain advantages, and the economics of running AI workloads at massive scale. Nvidia's roadmap from GB300 to the upcoming Vera Rubin line isn't merely iterative. It's designed to reset the cost and performance curves in ways that matter for both training and inference, giving Nvidia a structural edge that's difficult to dislodge with point optimizations

alone. A big part of that advantage stems from volume economics and supply chain positioning. Nvidia has locked in critical packaging capacity that remains a bottleneck for the broader industry, according to deep analysis from industry watchers, that translates into cost advantages that go beyond peak flops or benchmark scores, especially as demand for high bandwidth communication intensive workloads explodes. When you're building AI factories with tens of thousands of interconnected accelerators, real-world economics quickly trump theoretical spec sheets. Now toss Google's Gemini into the mix. There's no denying the model has generated buzz. Benchmark wins and growing share in certain usage patterns

have prompted some to declare the AI era rebalanced, but the underlying narrative that a single foundational model will suddenly flip the AI stack overlooks the reality that software ecosystems, developer momentum, and business models matter as much as raw model capability. Google's business remains deeply tethered to its advertising engine and significant shifts there would require not just technological superiority, but a reimagining of how search and commerce integrate in a post-LLLM world. What this means for Open AI is equally fascinating. Despite intense competition from Gemini to to specialist models from China and elsewhere, OpenAI's emphasis on trusted information, platform

integrations, and enterprise-ready tooling suggests that leadership and generative AI won't be decided solely by who wins a benchmark race next quarter. Competitive pressures are real. Indeed, OpenAI has declared multiple internal code-read initiatives this year in response to competitive threats, even ones we haven't heard about. But the next number. narrative that OpenAI is on the ropes is exaggerated. Taken together, this suggests an important framing shift. The AI era isn't about a single Silicon winner or a singular model, but about ecosystem strength, breadth of deployment, and economic scale. Invidia's mode isn't indestructible, but it's deeper and wider than most headlines concede. Meanwhile, the broader landscape

from Google to Open AI to emerging players will continue to evolve in ways that defy simplistic single-axis winner-take-all narratives. At least, according to Silicon angle. Nothing more for you today. Talk to you tomorrow. All. Pay off your home. Travel for life. Drive a Ferrari. In celebration of the world premiere of the Monopoly Big Board Buckslot machine by Aristocrat Gaming, Yamava Resort and Casino at San Manuel is giving one person a $1.6 million dream package. The biggest prize in Yamava's history. Club Serrano members can earn daily instant prizes and secure a spot in the finale May 29. Don't pass go and own it all. Only at Yamava, celebrating its 40th anniversary.

You win? Details at Yamava.com must be 21-20. Please gamble responsibly. Monopoly is a trademark of Hasbro. Hasbro is not a sponsor of this promotion.