koyfin.com > Glossary: [Excerpt] Buyback Yield % is the repurchase of outstanding shares over the existing market cap of a company. It is calculated using the formula:
Buyback Yield % = ((Issuance of Common Stock + Repurchase of Common Stock) / Latest Market Cap)
Google is using some of the 20B videos on YouTube to train its AI tool
cnbc.com: [Excerpt] Creators say they didn’t know Google uses YouTube to train AI.
Google is using its expansive library of YouTube videos to train its AI models, including Gemini and the Veo 3 video and audio generator, CNBC has learned.
The company confirmed to CNBC that it relies on its vault of YouTube videos to train its AI models, but Google said it only uses a subset of its videos for the training.
Such use of YouTube videos has the potential to lead to an intellectual property crisis for creators and media companies, experts said.
Google is using its expansive library of YouTube videos to train its artificial intelligence models, including Gemini and the Veo 3 video and audio generator, CNBC has learned.
The tech company is turning to its catalog of 20 billion YouTube videos to train these new-age AI tools, according to a person who was not authorized to speak publicly about the matter. Google confirmed to CNBC that it relies on its vault of YouTube videos to train its AI models, but the company said it only uses a subset of its videos for the training and that it honors specific agreements with creators and media companies.
@The_AI_Investor: Google is the only fully vertically integrated player.
⚡️
@ArtificialAnlys: Google is firing on all cylinders across AI - Gemini 2.5 Pro is equal #2 in intelligence, Veo 3 and Imagen 4 are amongst the leaders in media generation, and with TPUs they're the only vertically integrated player
Google is now equal #2 Artificial Analysis Intelligence Index with the recent release of the Gemini 2.5 Pro (June 2025) model, rivaling others including OpenAI, DeepSeek and Grok
Google Veo 3 now ranks second in the Artificial Analysis Video Arena Leaderboard only behind ByteDance’s new Seedance 1.0 model
Google Imagen 4 now occupies 2 out of the top 5 positions on the Artificial Analysis Image Arena Leaderboard
Google has a full stack AI offering with offerings across the application layer, models, cloud inference and hardware TPUs)
@ArtificialAnlys: Google has consistently been shipping intelligence increases in its Gemini Pro series
@ArtificialAnlys: Google Veo 3 now occupies second place in the Artificial Analysis Video Arena, after originally debuting in the first place. Still a significant leap over Google Veo 2!
@ArtificialAnlys: Google Imagen 4 Ultra Experimental and Imagen 4 Preview now rank 4th and 5th respectively in the Artificial Analysis Image Arena.
@ArtificialAnlys: Further analysis in our quarterly State of AI report (highlights report freely accessible on Artificial Analysis)
Over the past two weeks, data points highlighting the slowest paid click growth in Google’s history combined with the first month ever of declining web search volumes on Apple devices validate investors’ feeling – it’s not just be them changing their search behavior.
Growing pies and shrinking slices. There are two truths: (1) the “search” pie is growing, and (2) Google’s share is shrinking. The question that matters most is whether Google’s slice of the larger pie is still bigger than what their slice would have been pre Gen AI.
Is a growing pie commercial? Gen AI is adding queries we never asked Google before – create an image, draft an email response, and summarize topics. Most of these appear difficult to monetize, though Gen AI should also migrate more commercial queries online – if we can research better, we should convert at a higher rate.
Shrinking slices come in threes. AI user number updates from ChatGPT and MetaAI creep towards 1B, lowest ever paid click growth on Google, and Apple seeing first month decline in search volume. Combined, we estimate Google’s search share is closer to 65-70% vs. the 90% we often hear.
Google Search is now social. Investors scrutinize exposure growth to evaluate terminal risk/potential of social media rails – we did this with Meta and Snapchat in the face of TikTok download share gains. We should do the same with Google – Google is now firmly in this camp with revenue growth now viewed as a lagging indicator.
⚡️
@MonkEchevarria: Man, that Bernstein report on Google’s slipping search grip is a real eye-opener. I’m cautiously optimistic about GOOG, though. Sure, market share is down, and AI rivals are hungry, but their profitability is still a beast, ranking high against peers. Even with a pricey valuation dragging its value score, their cash flow machine might just keep ‘em ahead. Gotta watch that shrinking pie slice, tho; it’s a wild tech jungle out there.
🙂
☑️ #187 May 7, 2025
The $164 zone
@VolumeLeaders: $GOOG / $GOOGL - It would be disingenuous for me to say this bad news and subsequent drop was knowable in advance. But what I can demonstrate is that the level from which the decline began is a very strong s/r level that goes all the way back to last April. It's its 4th largest level since inception. VL provides this context by painting levels in which elevated institutional activity has occurred in the past.
Sept 2024 OPEX (and to a lesser degree April 2024 EOM) came in quite large and the level in which that inventory traded held as support on multiple tests from above going into EOY. When the 2025 decline began, it served as temp support before ultimately breaking lower.
Upon revisiting that level from below, it now appears to have held once more, just in the opposite direction.
The news itself wasn't foreseeable, but if news was going to come (positive or negative), this location is a pretty good candidate for where we'd see it. In other words, the $164 zone was a good place to reduce risk or trail stops if long given how much activity has transpired here in the past. Should that level be recaptured in the future, it may well serve as strong support once again. Time will tell.
@VolumeLeaders: $GOOG - Here's a closer look at the action since March. Several interactions with this level.
+ Related content:
sherwood.news: [Excerpt]Google tanks, dragging market lower, after Apple says it’s considering adding AI search in its browser.
APPLE’S CUE: SEARCHES IN BROWSER FELL FOR FIRST TIME IN APRIL
APPLE’S CUE: APPLE EXPLORING ADDING AI SEARCH TO ITS BROWSER
APPLE’S CUE: BELIEVES PERPLEXITY, ANTHROPIC COULD BE OPTIONS
justice.gov/atr > U.S. and Plaintiff States v. Google LLC [2020] (Remedies Hearing) (5/2/25): [Google Is a Monopolist in Ad Tech: May 2 hearing date to discuss potential remedies in the antitrust case] United States District Court for the Eastern District of Virginia Leonie Brinkema set Alphabet Inc.'s Google's newest antitrust trial related to its advertisement policy for September 22.
🙂
☑️ #186 May 1, 2025
Over 55M miles of fully autonomous driving data
@dmitri_dolgov: Over 55M miles of fully autonomous driving data shows the @Waymo Driver is making roads safer for everyone, especially for those most at risk – pedestrians, cyclists, and other vulnerable road users. Strong positive impact already and growing exponentially. waymo.com/blog/2025/05/waymo-making-streets-safer-for-vru
+ Related content:
waymo.com (5/1/25: [Excerpt] New Study: Waymo is reducing serious crashes and making streets safer for those most at risk.
The path to Vision Zero requires reducing severe crashes and improving the safety of those most at risk. Our latest research paper shows that the Waymo Driver is making significant strides in both areas. By reducing the most dangerous crashes and providing better protection for pedestrians, cyclists, and other vulnerable road users, Waymo is making streets safer in cities where it operates.
The paper, accepted to be published in the Traffic Injury Prevention Journal, expands on our Safety Impact Hub research, providing a deeper analysis of Waymo’s performance across 11 different crash types compared to human drivers. It also offers new insights into Waymo’s positive impact on serious injury crash rates.
Bill Ackman - Pershing Square Capital Management, L.P. (via Uncle Stock Notes): [Excerpt] Revenues grew at a robust 15% rate led by strength in its core advertising business and rapid deployment of Al innovations. In Search, the company has successfully rolled out Al-powered summary responses, called "Al Overviews". Early engagement with Al Overviews has been highly encouraging with consumers searching more frequently, with more detailed queries, and clicking through at higher rates. Google has been able to address two key investor concerns with introducing Al in Search around cost-to-serve and ad monetization. The company has reduced the compute cost per Al query by more than 90% over the last 18 months, a step-function cost reduction which has enabled Google to roll out Al Overviews broadly to more than a billion users. The company also recently introduced ads on Al Overviews and has explained that these ads are already monetizing at the same rate as regular Search ads. Over time, we expect more context-rich Al responses to drive even higher conversion rates and efficacy on ads.
Beyond Search, Google's Al capabilities continue to differentiate and strengthen the company's other business lines. Al-powered creative tools have propelled YouTube to be a leading destination for content creators and driven its success in the short-form video format and on Connected TVs, where YouTube is the number one streaming platform in the U.S. by watch time according to Nielsen. Google's Cloud segment, which is now a nearly $50 billion run-rate business, grew 31% in 2024 driven by demand for the company's enterprise Al offerings. Most notably, the segment's impressive growth rate was potentially capped by near-term capacity constraints, which the company is seeking to alleviate with its expanded $75 billion capital expenditure plan in 2025. With increasing scale, the Cloud segment has inflected from breakeven profitability in 2023 to 17.5% profit margins in the most recent quarter, with line of sight to achieving the greater than 30% profit margins enjoyed by peers.
More broadly, profitability across Alphabet is improving at a rapid clip. Consolidated profit grew 29% and profit margins expanded 370 basis points in 2024 (excluding one-time severance and restructuring charges) as cost control initiatives and headcount leverage are bearing fruit. Over the last two years, the company's total headcount has decreased 4% while revenue has grown 27%. In addition to driving revenue growth opportunities, Al integration is also enabling substantial automation gains in Google's internal workflows and operations. The company recently shared that more than a quarter of all new code is generated by Al and then reviewed and accepted by engineers. We expect these automation efficiencies to continue driving margin expansion over the long-term. To that end, we were encouraged to hear new CFO, Anat Ashkenazi, commit to building on the company's efficiency initiatives and evaluate where they might be able to accelerate their work.
Despite tremendous business momentum, Alphabet still trades at around 18x forward earnings, a highly discounted valuation for a business of its quality and growth prospects. As the company executes on its Al potential and margin expansion opportunity to drive rapid earnings growth, we expect its share price will increase to better reflect its strong underlying business fundamentals.
marketwatch.com (5/25/23): [Excerpt] ‘An unmatched business model’: Why Bill Ackman’s Pershing Square spent $1 billion on Alphabet stock.
Pershing disclosed earlier in May that it amassed a new position in Alphabet during the first quarter that amounted to 10.3 million shares across the Class A and Class C classes. That position was worth upwards of $1 billion as of the end of the quarter, though Ackman said Wednesday that Pershing’s average cost was about $94 to $94.50 a share.
☑️ #185 Apr 23, 2025 🟠 opinion
District Court Rules Google Is a Monopolist in Ad Tech
stblaw.com: [Excerpt] On April 17, 2025, U.S. District Judge Leonie Brinkema of the U.S. District Court for the Eastern District of Virginia issued a decision in the liability phase of the Google ad tech monopolization case brought by the Department of Justice (“DOJ”) and 17 states (collectively “Plaintiffs”). In an opinion that chronicled the history and evolution of web display advertising technology, the court found Google liable for monopolizing both the market for open-web display ad exchanges, which conduct the real-time auctions to place display advertisements on websites, and the market for ad servers used by website publishers, which manage the sale and display of advertisements on the publisher’s website. It also found Google liable for tying its products in the two markets together. The case will now move on to the remedy phase, with the parties due to submit their proposed schedules this week. Notably, the prayer for relief in the DOJ’s complaint indicates they planned to seek the divestiture of both DoubleClick and AdX.
justice.gov (4/17/25): [Excerpt] Department of Justice Prevails in Landmark Antitrust Case Against Google.
Today the Antitrust Division of the Department of Justice prevailed in its second monopolization case against Google. In United States et al. v. Google, the U.S. District Court for the Eastern District of Virginia held that Google violated antitrust law by monopolizing open-web digital advertising markets. According to the Court, Google “harmed Google’s publishing customers, the competitive process, and, ultimately, consumers of information on the open web.”
https://www.lexology.com: Click here to read the U.S. Federal District Court for the Eastern District of Virginia Court’s decision in U.S. v. Google.
justice.gov/atr > U.S. and Plaintiff States v. Google LLC [2020] (Remedies Hearing) [Excerpt] The U.S. filed suit to stop Google from protecting its search and advertising monopolies through a series of exclusionary agreements designed to lock up the primary avenues through which users access search engines. Trial began on September 12, 2023. Trial exhibits and documents are available on the case page.
Source: Antitrust Division - U.S. Department of Justice
usvgoogleads.com: [Excerpt] The antitrust trial of the United States v. Google, LLC. Google is accused of illegally monopolizing the digital advertising industry where they hold a mob-like control over who gets to earn revenue from ads, often depriving real journalism while allowing disinformation to profit.
Last August, the US Government ruled that Google illegally exercised monopoly powers by controlling which apps are on your phone, and worse, which apps are NOT on your phone.
Google controlled this through elaborate contracts with phone makers and carriers. The terms of the contracts threatened OEMs and carriers with wide ranging penalties for not satisfying vague or arbitrary reviews by Google.
This week, Google will be back in court with the Department of Justice to determine what the remedy for this should be.
Perplexity has been asked by both the DOJ and Google to provide input on this case. That’s rare. It tells you this moment matters—because what happens here affects the future of innovation, not just in search, but across the digital economy.
Ironwood: The first Google TPU for the age of inference
news.samsung.com: [Excerpt] Ironwood is our most powerful, capable and energy efficient TPU yet, designed to power thinking, inferential AI models at scale.
Ironwood is built to support this next phase of generative AI and its tremendous computational and communication requirements. It scales up to 9,216 liquid cooled chips linked with breakthrough Inter-Chip Interconnect (ICI) networking spanning nearly 10 MW. It is one of several new components of Google Cloud AI Hypercomputer architecture, which optimizes hardware and software together for the most demanding AI workloads. With Ironwood, developers can also leverage Google’s own Pathways software stack to reliably and easily harness the combined computing power of tens of thousands of Ironwood TPUs.
cloud.google.com: [Excerpt] Cloud Tensor Processing Units (TPUs). Accelerate AI development with Google Cloud TPUs.
Cloud TPUs optimize performance and cost for all AI workloads, from training to inference. Using world-class data center infrastructure, TPUs offer high reliability, availability, and security.
Announcing Ironwood, our most powerful, capable and energy efficient TPU yet, designed to power thinking, inferential Al models at scale.
🙂
☑️ #182 Apr 9, 2025
“Hey Ballie”
news.samsung.com: [Excerpt] Samsung and Google Cloud Expand Partnership, Bring Gemini to Ballie, a Home AI Companion Robot by Samsung.
Gemini on Google Cloud will enable Ballie to bring personalized interactions and proactive home assistance to users; available in United States and Korea this Summer.
samsung.com: Ballie is more than a home robot, it's your buddy.
Source: Samsung Electronics America, Inc.
🙂
☑️ #181 Mar 29, 2025 🟠 opinion
The Mediocrity of Modern Google
om.co: [Excerpt] The signs are clear for those willing to see them. Like Microsoft in the late ’90s and IBM before that, Google is proving ill-equipped to respond to fundamental shifts in user behavior. The company remains trapped in its “10 blue links” jail, even as the world moves toward direct, conversational AI interfaces. Despite helping pioneer the technologies that enabled this transition, Google seems unable to break free from its advertising-driven business model.
CME Group Will Introduce Tokenization Technology to Enhance Capital Market Efficiency Using Google Cloud's New Universal Ledger
investor.cmegroup.com: [Excerpt] CHICAGO and SUNNYVALE, Calif., March 25, 2025 /PRNewswire/ -- Further expanding their transformational partnership, CME Group and Google Cloud today announced that they are piloting solutions for seamless and secure wholesale payments and tokenization of assets. CME Group has successfully completed the first phase of integration and testing of Google Cloud Universal Ledger (GCUL), a novel distributed ledger from Google Cloud.
Google DeepMind and NVIDIA are working to build trust in generative AI through content transparency.
To power research and AI production efforts for its customers, Alphabet’s Google Cloud will be among the first to adopt the NVIDIA GB300 NVL72 rack-scale solution and NVIDIA RTX PRO™ 6000 Blackwell Server Edition GPU, also announced today at GTC.
NVIDIA will be the first industry partner to adopt SynthID, a Google DeepMind AI technology that embeds digital watermarks directly into AI-generated images, audio, text or video.
Intrinsic is an Alphabet company focused on making intelligently adaptive AI for robotics usable and valuable for manufacturers across industries. Today, the majority of the world’s installed industrial robots are manually programmed, with every movement hard-coded in a complex, expensive process.
Partnering with NVIDIA, the teams have built deeper and more intuitive developer workflows for Intrinsic Flowstate to support NVIDIA Isaac Manipulator foundation models for a universal robot grasping capability. Using foundation models for robotics will significantly reduce application development time and improve flexibility, with AI that can adapt effortlessly. At GTC, Intrinsic will also share an early OpenUSD framework streaming connection between Intrinsic Flowstate and NVIDIA Omniverse — enabling real-time visualization of robot workcells across platforms.
Silicon Valley riders: A 27-square-mile service area, part of Waymo’s broader plan to connect Silicon Valley and San Francisco
@Waymo: We’ve made millions of miles of memories over the past 15+ years, but today is special. We’re returning to where the journey began, gradually opening our doors to our first public riders in Mountain View, Los Altos, Palo Alto, and parts of Sunnyvale.
⚡️
@Waymo: We’ll gradually add riders to this new service area over time, as we work to bring Waymo One to more of the Bay Area. Keep an eye on the app if you’re an area local. Who’s ready to ride?
This rollout uses fully electric Jaguar I-Pace vehicles equipped with Waymo’s fifth-generation self-driving technology.
@tetumemo: [Translated from JP by Google] DeepSeek and o3-mini are hot topics, but Google's "DataGemma" is also pretty amazing.
In short, "The combination of RAG and RIG allows for accurate research with minimal hallucination (misinterpretation of facts)." What is RIG? What happens when you combine it with RAG? For details, please see the reply section ↓
+ Related content:
blog.google (9/12/24): [Excerpt] DataGemma: Using real-world data to address AI hallucinations.
DataGemma are the world’s first open models designed to help address the challenges of hallucination by grounding LLMs in the vast, real-world statistical data of Google's Data Commons.
How Data Commons can help tackle hallucination
As generative AI adoption is increasing, we’re aiming to ground those experiences by integrating Data Commons within Gemma, our family of lightweight, state-of-the art open models built from the same research and technology used to create the Gemini models. These DataGemma models are available to researchers and developers starting now.
DataGemma will expand the capabilities of Gemma models by harnessing the knowledge of Data Commons to enhance LLM factuality and reasoning using two distinct approaches:
RIG (Retrieval-Interleaved Generation) enhances the capabilities of our language model, Gemma 2, by proactively querying trusted sources and fact-checking against information in Data Commons. When DataGemma is prompted to generate a response, the model is programmed to identify instances of statistical data and retrieve the answer from Data Commons. While the RIG methodology is not new, its specific application within the DataGemma framework is unique.
Example query: ''Has the use of renewables increased in the world?'' applying DataGemma RIG methodology leverages Data Commons (DC) for authoritative data.
RAG (Retrieval-Augmented Generation) enables language models to incorporate relevant information beyond their training data, absorb more context, and enable more comprehensive and informative outputs. With DataGemma, this was made possible by leveraging Gemini 1.5 Pro’s long context window. DataGemma retrieves relevant contextual information from Data Commons before the model initiates response generation, thereby minimizing the risk of hallucinations and enhancing the accuracy of responses.
Example query: ''Has the use of renewables increased in the world?'' applying DataGemma RAG methodology showcases greater reasoning and inclusion of footnotes.
huggingface.co: [Excerpt] Finally, a Replacement for BERT.
Introduction
BERT was released in 2018 (millennia ago in AI-years!) and yet it’s still widely used today: in fact, it’s currently the second most downloaded model on the HuggingFace hub, with more than 68 million monthly downloads, only second to another encoder model fine-tuned for retrieval. That’s because its encoder-only architecture makes it ideal for the kinds of real-world problems that come up every day, like retrieval (such as for RAG), classification (such as content moderation), and entity extraction (such as for privacy and regulatory compliance).
Finally, 6 years later, we have a replacement! Today, we at Answer.AI and LightOn (and friends!) are releasing ModernBERT. ModernBERT is a new model series that is a Pareto improvement over BERT and its younger siblings across both speed and accuracy. This model takes dozens of advances from recent years of work on large language models (LLMs), and applies them to a BERT-style model, including updates to the architecture and the training process.
research.google and arxiv.org (10/11/18): [Abstract] BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding.
We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models, BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5% (7.7% point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement).
🙂
☑️ #172 Dec 17, 2024
Infosys strengthens strategic collaboration with Google Cloud to boost enterprise AI innovation, establishes Center of Excellence
infosys.com: [Excerpt] Leveraging Infosys Topaz and Google Cloud technology, the center of excellence will foster co-innovation to deliver transformative AI-powered solutions.
Infosys and Google Cloud share a strong legacy of delivering innovative data analytics and AI solutions, driving customer success. Infosys has already enabled more than 60,000 employees on Google Cloud. Combining the expertise of Infosys Topaz AI capabilities and Infosys Cobalt cloud capabilities, with Google Cloud's gen AI technology, the center of excellence will help address complex challenges across both business and technology domains using innovative enterprise AI solutions. The collaboration will enable a conducive environment for enterprises to co-create customized solutions in areas such as contact center AI, software development lifecycle, agentic AI, speech-to-speech, text-to-image, and application modernization, aimed at optimizing efficiency and business performance.
developers.googleblog.com (12/11/24): [Excerpt] The next chapter of the Gemini era for developers. Build with Gemini 2.0 Flash.
Building on the success of Gemini 1.5 Flash, Flash 2.0 is twice as fast as 1.5 Pro while achieving stronger performance, includes new multimodal outputs, and comes with native tool use. We’re also introducing a Multimodal Live API for building dynamic applications with real-time audio and video streaming.
Starting today, developers can test and explore Gemini 2.0 Flash via the Gemini API in Google AI Studio and Vertex AI during its experimental phase, with general availability coming early next year.
🙂
☑️ #170 Dec 13, 2024
Unlock the Infinite Possibilities of XR With Galaxy AI
news.samsung.com: [Excerpt] Available for both headsets and glasses, this new platform leverages the power of Gemini by bringing together a conversational user interface (UI) and a contextual understanding of the world around you. It offers cutting-edge capabilities beyond gestures or a controller, utilizing voice and natural conversation to elevate the experience. With these accurate, personalized interactions, the platform will be your helpful AI assistant.
Code named “Project Moohan”, the first headset designed for Android XR is poised to bring this experience to life in the near future. The name “Moohan”, meaning ‘infinity’ in Korean, connotes our belief in delivering unparalleled, immersive experiences within an infinite space. Equipped with state-of-the-art displays, Passthrough capabilities and natural multi-modal input, this headset will be your spatial canvas to explore the world through Google Maps, enjoy a sports match on YouTube or plan trips with the help of Gemini. All these experiences come with lightweight, ergonomically optimized hardware designed to ensure maximum comfort during use.
Intersect Power forms strategic partnership with Google and TPG Rise Climate to co-locate data center load and clean power generation
intersectpower.com: [Excerpt] Collaboration accelerates the evolution of data center development and power provision, reducing costs, increasing reliability, and supporting local economic development. New ~$800M funding round led by TPG and Google will accelerate partnership’s development.
Under the terms of the partnership, Intersect Power will build new clean energy assets, with Google providing offtake via newly constructed data center campuses as an anchor tenant in co-located industrial parks. Once built, the Google data center would come online alongside its own clean power, bringing new generation capacity to the grid to meet its own load.
To further its vision throughout the U.S., Intersect Power also announced a more than $800 million funding round led by TPG Rise Climate and Google, with participation from Climate Adaptive Infrastructure and Greenbelt Capital Partners. Morgan Stanley & Co. LLC acted as financial advisor on the transaction.
TPG Rise Climate is the dedicated climate investing strategy of TPG’s $25 billion global impact investing platform. TPG Rise Climate pursues climate-related investments that benefit from the diverse skills of TPG’s investing professionals, the strategic relationships developed across TPG’s existing portfolio of climate-focused companies, and a global network of executives and advisors. The fund takes a broad-based sector approach to investment types, from growth equity to buyouts, and focuses on climate solutions in the following thematic areas: clean electrons, clean molecules and materials, and negative emissions. Jim Coulter, TPG Founding Partner and Executive Chairman, serves as a Managing Partner of TPG Rise Climate. Former U.S. Treasury Secretary Hank Paulson serves as TPG Rise Climate’s Executive Chairman. For more information, please visit therisefund.com/tpgriseclimate
🙂
☑️ #168 Dec 9, 2024
AI computing capacity fore leading tech companies
@thexcapitalist: $GOOG is way ahead in AI race than you think.
It has the largest AI computing capacity among cloud providers.
What’s even better is that 70% of its capacity comes from its own TPU chips.
$MSFT and $AMZN are completely dependent on Nvidia.
@sundarpichai: Introducing Willow, our new state-of-the-art quantum computing chip with a breakthrough that can reduce errors exponentially as we scale up using more qubits, cracking a 30-year challenge in the field. In benchmark tests, Willow solved a standard computation in <5 mins that would take a leading supercomputer over 10^25 years, far beyond the age of the universe(!).
⚡️
@sundarpichai: We see Willow as an important step in our journey to build a useful quantum computer with practical applications in areas like drug discovery, fusion energy, battery design + more. Details here:
@joecarlsonshow: So many people believe Waymo can't scale. Here are the numbers:
50,000 rides a week 6 months ago.
100,000 rides a week 3 months ago.
150,000 rides a week 1 months ago
175,000 rides last week.
Waymo is scaling quickly. And they have good reason to become profitable over time. Waymo vehicles are cheaper every iteration, like the Waymo 6th generation, which is "optimized for costs" and is significantly cheaper than previous versions, despite being able to operate in harsher conditions.
The central costs of having support staff and ride monitoring should scale efficiently as the number of rides grows (as it's done with Uber).
Large capex investments into building and buying vehicles up-front should moderate over time. Lidar is becoming cheaper over time as well.
Waymo will get more experience and better and faster at opening in new cities as time goes on.
Google took a different business (Cloud) from deeply in the red to a profitable business over time. They know how to increase margins of capex heavy businesses.
If you're on the "they can't scale" side, please explain why.
Our previous road trips to the Sunshine State’s challenging rainy conditions have been invaluable in advancing our autonomous driving capabilities and we’re honored to return and bring our service – which already provides over 150,000 trips per week across Phoenix, Los Angeles, San Francisco, and Austin – to the people of Miami.
We'll first collaborate with our fleet partner Moove in Phoenix, where they'll begin taking on the management of our fleet operations, facilities, and charging infrastructure. This strategic partnership is designed to maintain and enhance the clean, consistent experience that our riders have come to expect and rely on. In both Phoenix and eventually Miami, Waymo will continue to offer our service through the Waymo One app, and remain responsible for validation and operation of the Waymo Driver.
🙂
☑️ #165 Dec 4, 2024
Google Cloud Lands Partnership with Air France-KLM to Transform its Data and Generative AI Strategy
prnewswire.com : [Excerpt] The global airline group will upgrade the value of its data and get its AI & generative AI ready for take off with Google Cloud's data, analytics and AI tools and expertise.
Air France-KLM has engaged in a major transformation of its three legacy data centers toward a multi-cloud strategy. Moving applications to the cloud and accelerating data value and insights, Air France-KLM will leverage Google Cloud's data and analytics tools, including BigQuery. With Google Cloud's commitment to security and privacy, Air France-KLM will maintain full ownership and control of its data, while improving time-to-market, innovation, and less data processing latency to better serve its business units in their digital journeys.
Through this work, the group will gain a deeper understanding of passenger preferences, travel patterns, and behaviors to offer newly tailored travel options and services, and be able to enhance flight, airport, and business operations, with the time needed for data analysis in predictive plane maintenance already dropping from hours to minutes, improving efficiency and reducing environmental impact.
Google plans "historic" data centre expansion in Finland
yle.fi: [Excerpt] The expansion will involve the purchase of parcels of land in the municipalities of Muhos and Kajaani, which the government gave wildlife Agency Metsähallitus a green light to approve on Thursday.
The acquisition of the lands — covering some 1,400 hectares — will cost the company an estimated 27 millions euros, a government press release said, while the expansion of the data centre network in Finland could potentially create hundreds of jobs in the Muhos and Kajaani region.
Google already operates a large data centre in Hamina, southeast Finland.
blog.google (5/20/24): Our first offsite heat recovery project lands in Finland. The project will provide 80% of heat demand for Finnish neighborhoods.
🙂
☑️ #163 Nov 26, 2024
Sum of the Parts. Alphabet's pieces could be worth $260 per share, more than 50% higher than Thursday's close.
Google Announces Australia Connect Initiative to Build Bosun Subsea Cable Connecting Darwin to Christmas Island
submarinenetworks.com: [Excerpt] Google is collaborating with several key partners for the Australia Connect initiative, including: NEXTDC, SUBCO, Vocus, along with state and local governments in Darwin, Perth, and the Sunshine Coast.
NEXTDC is a Data Centre-as-a-Service provider in Australia. NEXTDC owns and operates the Sunshine Coast Cable Landing Station (NEXTDC SC1) in Maroochydore, Sunshine Coast, where hosts JGA South and Google's Tabua cable systems. NEXTDC will work in partnership with Google to establish cable landing stations in Darwin and Melbourne as well.
SUBCO and Google will collaborate on building coordinated cable landing infrastructure in Maroubra, NSW, and Torquay, Victoria as well as new infrastructure that connects these locations back to the respective parties’ cable landing stations. This infrastructure will support SUBCO’s transcontinental submarine cable system connecting Sydney, Melbourne, Adelaide and Perth (SMAP) and Google’s upcoming transpacific cable systems, Tabua and Honomoana. SUBCO is building the SMAP cable system connecting Sydney, Melbourne (Torquay), Adelaide and Perth.
vocus.com.au: [Excerpt] Vocus partners with Google on Australia Connect, linking Southeast Asia and North America via Australia. Australia Connect will extend beyond the previously announced Pacific Connect system that links Australia and the West Coast of the United States with a multi-terabit ring network through and between Fiji and French Polynesia, landing in Sydney and Melbourne.
cloud.google.com: [Excerpt] Australia Connect initiative delivers new digital pathways for the Indo-Pacific. Google is deeply committed to building a strong digital future for all Australians. In 2021, we launched the Digital Future Initiative, a $1 billion AU$, five-year initiative that builds on our long-term commitment to infrastructure, local partnerships, and research capabilities. Analysys Mason estimates Google’s previous submarine cable deployments in Australia will lead to a cumulative increase in GDP of $98.5 billion AU$ ($64 billion US$) between 2022 and 2026 and support the creation of around 68,000 additional jobs by 2027.
cloud.google.com: (10/25/23): [Excerpt] Connecting the South Pacific with new subsea cables. In collaboration with a number of partners, including Fiji International Telecommunications, Office of Posts and Telecommunications of French Polynesia, APTelecom and Vocus Group, the South Pacific Connect initiative will deliver new international subsea cables connecting Fiji and French Polynesia to both the United States and Australia.
Source: Google LLC
cloud.google.com (5/23/24): [Excerpt] Improving connectivity and accelerating economic growth across Africa with new investments. Anchored in Kenya, the Umoja cable route will pass through Uganda, Rwanda, Democratic Republic of the Congo, Zambia, Zimbabwe, and South Africa, including the Google Cloud region, before crossing the Indian Ocean to Australia. Umoja’s terrestrial path was built in collaboration with Liquid Intelligent Technologies to form a highly scalable route through Africa, including access points that will allow other countries to take advantage of the network.
spectrum.ieee.org: [Excerpt] For a decade, Google Research has helped advance nuclear energy R&D.
The proposed plan for the nuclear energy R&D group (affectionately known as NERD) was based on input from similarly minded colleagues. The problems we could address were determined by who we could work with externally, as well as Google’s usual strengths: people, tools, capabilities, and reputation. I proposed a three-pronged effort consisting of immediately impactful fusion research, a long shot focusing on an “out there” goal, and innovation advocacy in Washington, D.C. Some years later, we added sponsored research into the cutting-edge field of nuclear excitation. The NERD effort, started 10 years ago, is still bearing fruit today.
These programs all came from a question that I asked anybody who would listen: What can Google do to accelerate the future of nuclear energy?
A technician works on the TAE fusion machine known as “Norman” (c-2W). Source: TAE Technologies, Inc.
Norman runs up to 1,000 experiments per month – experiments that are all developed and executed from our command center, a.k.a our control room.
On any given day, a concert of TAE engineers, physicists, diagnostics specialists and technicians converge in this area to orchestrate the next sequence of tests, monitor and evaluate Norman’s plasma performance, and incorporate AI and machine learning to rapidly advance their findings.
“We have one of the most advanced machine learning and AI-enhanced operations of any machine in the fusion space that I’m familiar with,” Binderbauer said on the tour. “There’s a long-standing history between TAE and Google‘s inner research team working on machine learning and other tools to combine the spontaneity and creativity of the human mind and the untiring calculating ability of the computer and its infinite depth of memory. Every prior experiment ever run on Norman or predecessor machines or anything else … is available to the artificial brain, and so it can simulate all that historical learning and bring that to the decision process for each experiment. What we’ve learned is that by combining those two things, you make exponentially the biggest progress.”
Source: TAE Technologies, Inc.
🙂
☑️ #160 Nov 21, 2024
PwC opens Google Cloud AI Experience Zones to feature solutions powered by Google Cloud’s Generative AI
pwc.com: [Excerpts] November 21, 2024 – PwC today announced the opening of its first Google Cloud AI Experience Zone at the firm’s US Acceleration Center in Bengaluru, India. It also announced additional Google Cloud AI Experience Zones are coming to Boston, MA and San Francisco, CA. At each location, PwC will invite clients to get hands-on experience with business solutions that are powered by Google Cloud’s leading AI technology, including Gemini models, Vertex AI and other AI services.
PwC’s leading industry capabilities enable business model reinvention and help companies transform their business, enhance productivity and maintain regulatory compliance. These Google Cloud AI Experience Zones give customers a space to engage with AI agents and see in-person how they can assist humans in solving industry challenges. PwC’s suite of offerings spans various industries and includes:
Health AI industry solutions
Generative AI (GenAI) capabilities powered by our GenAI Agent & Platforms
Data marketplace capabilities
Retail experience capabilities powered by Retail Search & GenAI
Google has manipulated its control of Chrome and Android to benefit itself
U.S. and Plaintiff States v. Google LLC [2020]
courtlistener.com: [Excerpt] NOTICE of Plaintiffs' Initial Proposed Final Judgment by UNITED STATES OF AMERICA (Attachments: # 1 Text of Proposed Order)(Herrmann, Karl) (Entered: 11/20/2024)
For more than a decade, Google has unlawfully maintained its monopolies in general search services and search text advertising through a web of anticompetitive practices. As this Court found after a lengthy trial, “Google is a monopolist, and it has acted as one to maintain its monopoly” over both the general search services and search text advertising markets. See Mem. Op., United States et al. v. Google LLC, 20-cv-3010 (APM), ECF No. 1032 (“Op.”), at 4. Google has manipulated its control of Chrome and Android to benefit itself, while sharing monopoly profits under conditions to induce third parties across the ecosystem to help Google maintain its monopolies. Google’s exclusionary conduct has, among other things, made Google the near-universal default for search and ensured that virtually all search access points route users’ valuable queries and interaction data to Google. Google’s unlawful behavior has deprived rivals not only of critical distribution channels but also distribution partners who could otherwise enable entry into these markets by competitors in new and innovative ways. Google’s conduct has resulted in significant anticompetitive effects—causing “market foreclosure,” “preventing rivals from achieving scale,” and “diminishing the incentives of rivals to invest and innovate.” Op. at 216.
justice.gov: U.S. and Plaintiff States v. Google LLC [2020]
🙂
☑️ #158 Nov 15, 2024
NVIDIA Accelerates Google Quantum AI Processor Design With Simulation of Quantum Device Physics
nvidianews.nvidia.com: [Excerpt] NVIDIA CUDA-Q Platform Enables Google Quantum AI Researchers to Create Massive Digital Model of Its Quantum Computer to Solve Design Challenges.
Understanding noise in quantum hardware designs requires complex dynamical simulations capable of fully capturing how qubits within a quantum processor interact with their environment.
These simulations have traditionally been prohibitively computationally expensive to pursue. Using the CUDA-Q platform, however, Google can employ 1,024 NVIDIA H100 Tensor Core GPUs at the NVIDIA Eos supercomputer to perform one of the world’s largest and fastest dynamical simulation of quantum devices — at a fraction of the cost.
People are changing the world with #generativeAI, from drug discovery to chatbots to autonomous machines and beyond. To achieve these breakthroughs, they need an AI factory - a purpose-built AI engine that's always available and can help ramp their capacity to build AI models at scale. Enter Eos. An extremely large-scale NVIDIA DGX SuperPOD, Eos is where NVIDIA developers create their leading-edge AI innovations using accelerated computing infrastructure and fully optimized software. Eos is built with 576 #NVIDIADGX H100 systems, NVIDIA Quantum-2 InfiniBand networking and software, providing 18.4 exaflops of AI performance and featuring a total of 4,608 H100 GPUs. Ranked #9 in the TOP500 list of fastest supercomputers, Eos is the apotheosis of NVIDIA's ongoing commitment to pushing the boundaries of AI technology and infrastructure. Learn more: https://nvda.ws/3wegRbl#datacenter
🙂
☑️ #158 Nov 4, 2024
OpenAI's ChatGPT Search takes on Google
AI Supremacy: The Future of AI is expensive and all about scale, Gen AI search enters pivotal 2025 period. Read to the end for AI Tools.
earning.google: [Excerpt] Grasp new topics and deepen your understanding with a 💬 conversational learning companion that adapts to your unique curiosity and learning goals.
Over a quarter of new code at Google is AI-generated, then reviewed and adopted by engineers
AI Disruption: During Google’s Q3 2024 earnings call, CEO Sundar Pichai stated: “Over a quarter of new code at Google is AI-generated, then reviewed and adopted by engineers.”
However, Google engineers were quick to challenge this claim.
Google Engineer 1: asdfman123
I work at Google and just finished my day. My work could technically be counted as "AI-generated code," but the code completion engine is mostly good at completing lines of code I’ve already half-written. For example, if I type “function getAc...,” it auto-completes to “function getActionHandler()” and might suggest suitable parameters and a decent doc comment.
Overall, it’s a useful productivity tool, but it doesn’t involve actual engineering work. It’s roughly as effective as Copilot, maybe even slightly less (though I haven’t used Copilot recently).
Google Engineer 2: NotAnOtter
I also worked at Google (until last Friday). I completely agree with you. My take is:
These claims are obviously exaggerated; they may even be counting fully automated code checks/pull requests (CL/PR) as “AI-generated,” which have been around for over a decade.
I previously mentioned that if a 10-person team’s productivity matches that of an 8-person team using something like Copilot, it’s fair to say “AI replaced two engineers.” More importantly, if this were true, tech leaders would have touted it long ago. Copilot and similar tools have existed long enough, yet no one claims “we’ve replaced X% of our workforce with AI.” So, following a “denial of the consequent” logic, I’d argue that using Copilot hasn’t fundamentally sped up development
@googlecloud: Live from the App Dev & Infrastructure #GoogleCloudSummit: Our AI Hypercomputer now includes Trillium TPUs in preview, Axion-powered C4A VMs in GA, and, coming soon, A3 Ultra VMs powered by NVIDIA H200 GPUs!
cloud.google.com: Powerful infrastructure innovations for your AI-first future.
@Google (11/1/24): At #GoogleIO in May, we announced Trillium, the sixth generation of our custom-designed chip known as the Tensor Processing Unit, or TPU. But what exactly *is* a TPU, and how is it different from a CPU or GPU? A Google expert explains ↓
blog.google (10/30/24): [Excerpt] Ask a Techspert: What's the difference between a CPU, GPU and TPU?
Source: Google LLC
🙂
☑️ #153 Oct 29, 2024
The best inventions of 2024: Google NotebookLM & Google DeepMind AlphaFold 3
time.com: [Excerpt] Released in July 2023 by Google Labs, the AI-powered NotebookLM helps users make sense of complex information by digesting up to 50 sources—documents, YouTube videos, even entire books—and summarizing them, suggesting follow-up questions, or creating a study guide. A feature released in September that converts such source information into an eerily realistic podcast, complete with AI-generated hosts, went viral. Steven Johnson, Google Lab’s editorial director and a bestselling nonfiction author, says he long dreamed of a digital companion that could take great amounts of information about a subject and help you make sense of it all. “I really spent 30 years effectively fantasizing about what eventually became NotebookLM,” he says.
time.com: [Excerpt] In 2020, Google DeepMind unveiled AlphaFold 2, a machine-learning algorithm that in 2022 would prove capable of predicting, with near-experimental accuracy, the structure of nearly every known protein. The breakthrough recently earned co-founder and CEO Demis Hassabis and AlphaFold team lead John Jumper a Nobel prize in chemistry. Now there’s a sequel: AlphaFold 3, released in May, predicts how proteins interact with DNA and other molecules. “AlphaFold 2 was a story about proteins,” Jumper says. “AlphaFold 3 is about how proteins talk to everything else.” Many in pharmaceuticals and medicine believe that the new model’s predictions will accelerate research and drive faster drug discovery
Russia Fines Google $2.5 Decillion Over YouTube Bans - RBC
rbc.ru: [Excerpts] [Translation] The requirements of Russian TV channels to Google reached ₽2 undecillions. Is there any chance to recover such an amount?
The amount of claims of Russian TV channels to Google due to the blocking of accounts on YouTube has increased to ₽2 undecillion. The court ordered the company to restore access, and until it does it, the penalty is growing every day.
The Russian court imposed the specified fine on Google due to the fact that the company did not comply with the requirement to restore Russian media accounts on YouTube (video hosting belongs to Google). The victims as third parties in the case are TV channels "Zvezda", "Channel One", VGTRK (TV channels "Russia 1", "Russia 24", etc.), "Parliamentary Television", "Moscow Media", "TV-Center", "GPM Entertainment Television", "Public Television of Russia", "TV Channel 360", "TRK Petersburg", "Orthodox Television Foundation", "National Sports TV Channel", "Technological Company Center", as well as IE Simonyan M.S., representing the YouTube channel of TV presenter Margarita Simonyan.
@AndrewCurran_: Sundar Pichai said on the earnings call today that more than 25% of all new code at Google is now generated by AI
☑️ #153 Oct 23, 2024 🟠 opinion
Breaking Up Google Is a Fool’s Game
wsj.com: [Excerpt] Trustbusters intent on sparking innovation and competition risk harming it, to Beijing’s benefit.
A word to the Justice Department, which is considering whether to push for a breakup of Google to spur competition in the online search market: Don’t. We’ve seen this movie before and it ended poorly. The 1984 breakup of AT&T was supposed to ignite competition and innovation but did the opposite. If we rush to dismantle Google, we risk undermining our global competitiveness against China and other adversaries.
Honeywell And Google Cloud To Accelerate Autonomous Operations With AI Agents For The Industrial Sector
honeywell.com: [Excerpt] Google Cloud AI to enhance Honeywell’s product offerings and help upskill the industrial workforce. New solutions will connect to enterprise-wide industrial data from Honeywell Forge, a leading IoT platform for industrials.
This partnership will bring together the multimodality and natural language capabilities of Gemini on Vertex Al - Google Cloud's Al platform - and the massive data set on Honeywell Forge, a leading Internet of Things (loT) platform for industrials. This will unleash easy-to-understand, enterprise-wide insights across a multitude of use cases. Honeywell's customers across the industrial sector will benefit from opportunities to reduce maintenance costs, increase operational productivity and upskill employees. The first solutions built with Google Cloud Al will be available to Honeywell's customers in 2025.
Google and Kairos Power Partner to Deploy 500 MW of Clean Electricity Generation
kairospower.com: [Excerpt] Google and Kairos Power partner to deploy 500 MW of clean electricity generation.
Under the agreement, Kairos Power will develop, construct, and operate a series of advanced reactor plants and sell energy, ancillary services, and environmental attributes to Google under Power Purchase Agreements (PPAs). Plants will be sited in relevant service territories to supply clean electricity to Google data centers, with the first deployment by 2030 to support Google’s 24/7 carbon-free energy and net zero goals.
blog.google: [Excerpt] New nuclear clean energy agreement with Kairos Power. By procuring electricity from multiple reactors — what experts call an “orderbook” of reactors — we will help accelerate the repeated reactor deployments that are needed to lower costs and bring Kairos Power’s technology to market more quickly. This is an important part of our approach to scale the benefits of advanced technologies to more people and communities, and builds on our previous efforts.
The Engineering Test Unit is Kairos Power’s largest-scale hardware demonstration to date and the first that tests the integration of key systems, structures and components that will go into the Hermes demonstration reactor. This non-nuclear iteration is helping to deliver technology, licensing, supply chain and build certainty on our path to achieving cost certainty for KP-FHR technology. Hear how ETU is driving innovation forward to advance Kairos Power’s clean energy mission in this video with co-founders, Mike Laufer, Ed Blandford, and Per Peterson
@Exponential View (update; 10/21/24): The key point to take away is that Google’s bulk order will put Kairos on an appealing learning curve.
