J Keynes Profile picture
Oct 20 20 tweets 16 min read Read on X
$SERV 🔥Serve Robotics: The AI Robotics Platform Powering the Future of Drones, Couriers, Logistics, and Sidewalk Autonomy🔥

The next addition to my Speculative AI portfolio is $SERV, Serve Robotics. So far I’ve added $ASAN, $RXRX, and $PATH, each representing a different layer of applied AI. Now I’m adding Serve because the next frontier isn’t digital AI in the cloud, it’s physical, ROBOTIC AI in the real world. And I'm feeling SUPER bullish!

Serve builds autonomous, AI-powered delivery robots that move at walking speed through real city environments. Each robot runs NVIDIA Jetson Orin edge compute fused with Ouster LiDAR, operating at Level 4 autonomy. $NVDA NVIDIA is not just a technology supplier, it is also an investor and strategic partner that provided capital and simulation tools that help train Serve’s autonomy stack.

The company spun out of $UBER Uber in 2021, with Uber still the largest shareholder and commercial partner. The robots already run in Los Angeles, Miami, Dallas, Atlanta, and Chicago, integrated directly into Uber Eats, logging over 100,000 deliveries with up to 99.8 percent reliability.

I’m adding it because AI-powered robotics is about to explode, and it’s almost impossible to find a small-cap name that combines real technical leadership, proven deployments, and a team that understands the bigger picture.

That's why I like Serve: $SERV Serve isn’t just delivering food; it’s building an autonomy OS that could power sidewalk robots, drones, warehouse movers, sanitation robots, or any form of short-range logistics. With partners like Uber, NVIDIA, Vinod Khosla, and Magna International, it’s positioned to become the platform layer for urban robotics. Let's dive in 🧵Image
$SERV The story behind $SERV shows how real this platform already is.

Serve started inside Postmates as its internal robotics division, Postmates X. When $UBER Uber acquired Postmates in 2020, CEO Ali Kashani negotiated to spin the robotics unit out as an independent company in 2021. Uber remained a major shareholder and became Serve’s first commercial partner, integrating its autonomous sidewalk robots directly into Uber Eats.

The spin-off gave Serve freedom to partner beyond Uber, opening doors to trials with DoorDash, Walmart, and 7-Eleven. It also kept Uber financially aligned while giving Serve full control of its roadmap, autonomy in both name and structure.

Then came NVIDIA. In 2023 filings with the SEC, Serve disclosed that NVIDIA had taken a multi-million-share equity stake, about 3.6 million shares, and received convertible notes that were later exchanged for additional stock. NVIDIA provided not only capital but also technical support through its Jetson Orin edge platform and Isaac robotics simulation stack. This partnership helped Serve reach Level 4 autonomy in real-world environments.

By late 2024, NVIDIA had exited its stake, locking in early gains, but the collaboration had already given Serve the compute architecture and AI infrastructure it still runs today.

Uber gave Serve the demand and distribution engine. NVIDIA gave it the brain. Together they accelerated Serve’s leap from concept to commercial autonomy, a rare feat for any small-cap robotics company.Image
$NVDA NVIDIA just featured $SERV as a showcase for physical AI on its own website.

In NVIDIA’s article “Serve Robotics Is Ushering Food Delivery Into the Mainstream,” the company highlights how Serve’s delivery robots rely on three NVIDIA GPUs, including one Orin AGX, for real-time processing of camera and LiDAR data during navigation through city streets.

NVIDIA notes that Serve’s autonomy depends on its Isaac SDK and Isaac Sim, which allow the company to model real-world edge cases and retrain its AI in simulation before redeployment. CEO Ali Kashani explains, “When you’re out in the real world and you see those edge cases, it’s really valuable to model those and then try them again in simulation. A lot of testing and validation happens in Isaac Sim.”

The article also mentions Serve’s acquisition of Vayu Robotics, integrating foundation-model training with simulation-based learning to make its robots safer and more intelligent.

When NVIDIA publicly writes about a partner this way, it signals real validation. Serve’s fleet doesn’t just use NVIDIA hardware; it helps define how NVIDIA’s robotics ecosystem performs in the real world.Image
$SERV $UBER Uber still owns nearly 5 million shares of $SERV and its vision for the company is clear.

According to Uber’s latest 13F filing, the company holds roughly 4.9 million shares of Serve Robotics, maintaining a major minority stake after spinning the division out in 2021. Uber did not just incubate Serve; it remains deeply invested in its success because robotic delivery fits perfectly into Uber’s long-term logistics vision.

Serve’s robots already complete deliveries for Uber Eats in Los Angeles, Miami, and other markets. The companies have a multi-year agreement to deploy up to 2,000 autonomous robots, part of Uber’s plan to expand delivery without scaling labor or emissions.

Uber has described autonomous delivery as a cornerstone of its future logistics stack, pairing AI dispatch with zero-emission mobility to cut costs and improve urban efficiency. Uber’s continued ownership and partnership serve as powerful validation from one of the most proven innovators in modern logistics technology.

It is not just NVIDIA that sees Serve as a crucial part of its early AI robotics ecosystem. Uber’s investment and integration show that two of the world’s most advanced technology companies are betting on Serve’s platform as the bridge between artificial intelligence and real-world autonomy.Image
$SERV What $SERV’s delivery robots actually do every day

Serve’s fleet operates in dense city environments across Los Angeles, Miami, Dallas, Atlanta, and soon Chicago. Each robot carries up to 15 gallons of cargo, enough for four 16-inch pizzas, and runs up to 14 hours per charge across 48 miles of range.

They navigate sidewalks and crosswalks at up to 11 mph, handling real weather, curbs, pedestrians, and even construction zones. Using Ouster REV7 LiDAR, high-resolution cameras, and redundant sensors, the robots perceive depth and motion like an autonomous vehicle but at human scale. The result: Level 4 autonomy, where no human driver or remote operator is required within the robot’s operating domain.

Serve Robotics isn’t testing theory. It’s one of the first autonomous vehicle companies to commercialize Level 4 autonomy for delivery robots.

That means no human is in the loop for safety within the robot’s defined operational domain. Every delivery is handled by the robot itself, using redundant sensors, NVIDIA Jetson Orin compute, and edge AI for navigation, braking, and obstacle avoidance.

Level 2 and 3 systems still depend on remote human control, which limits safety and scalability. Serve’s Level 4 architecture eliminates that dependency, improving economics and reliability.

The company’s deep-tech moat comes from this autonomy stack, which sits between remote-controlled robotics and fully self-driving (Level 5) systems, delivering real-world scalability now while others are still in R&D and face regulatory uncertainty.

Serve isn’t promising autonomy. It’s already shipping it.Image
$SERV Image
$SERV Why the Vayu Robotics integration is a huge deal for Serve

Serve’s recent acquisition of Vayu Robotics combines two of the hardest things in autonomy: massive real-world data and foundation-model intelligence.

Vayu’s system trains large AI models that learn navigation directly from multimodal data including LiDAR, cameras, and simulation rather than relying on hand-coded rules. This allows robots to adapt to new cities, lighting, terrain, and unpredictable human behavior with minimal retraining.

When fused with Serve’s unmatched sidewalk dataset, it creates an AI loop that can learn from every delivery. The models are refined through NVIDIA Isaac Sim and redeployed to the fleet, improving safety, efficiency, and speed with each update.

The result is a platform that scales like software while moving through the physical world. It is the technical leap that pushes Serve toward the frontier of “physical AI.”

Vayu + Serve is a huge AI data/training robotics flywheel for the future!Image
$SERV The Vinod Khosla connection takes $SERV to another level

Vinod Khosla, founder of Sun Microsystems and one of the most influential investors in AI and deep technology, joined Serve Robotics’ Advisory Board through the Vayu Robotics acquisition. He also took a direct financial position, receiving Serve stock and warrants tied to autonomy performance milestones and holding a stake via Khosla Ventures, which funded Vayu.

Khosla’s reputation carries enormous weight. He was an early backer of OpenAI, $QS QuantumScape, and $DASH DoorDash, and has spent decades identifying technology inflection points that change industries. His endorsement of Serve is not passive; it signals that he views foundation-model robotics as the next trillion-dollar platform shift.

Khosla said, “AI models are driving a new class of robotics across a range of industries. We invested early in Vayu because last-mile delivery stood out as one of the applications where autonomous delivery robots could create immense value.”

Readers of my account know I follow him closely. He is one of the sharpest minds in technology, the kind of investor who shaped how the internet, renewable energy, and now AI infrastructure evolved. His decision to take equity and a board role in Serve Robotics is a massive vote of confidence in the company’s mission to scale real-world autonomy.Image
$SERV $DASH $QS Image
$SERV Could Serve have a more visionary backer on its board? NO! Image
$SERV $DASH The DoorDash partnership is massive validation for $SERV

Vinod Khosla knows DoorDash well. His firm backed the company in its early days and helped it scale into the number-one food delivery platform in America. Now the company he advises and holds equity in, Serve Robotics, is officially partnering with DoorDash on autonomous deliveries across the United States.

Serve will fulfill DoorDash orders under a multi-year agreement, starting in Los Angeles and expanding nationwide. The robots already operate in five major cities including Los Angeles, Miami, Dallas, Chicago, and Atlanta, and have completed tens of thousands of real deliveries.

What makes this deal so important is that Serve’s partnerships are not exclusive. The same fleet that runs Uber Eats orders can also handle DoorDash deliveries, similar to how human couriers switch between apps. That means higher robot utilization, fewer empty miles, and stronger economics.

Serve CEO Ali Kashani called the move “a good validation because now both of the major platforms in the country are investing heavily in autonomy.” For years, Serve has argued that sidewalk robots would become a core layer of logistics. DoorDash agreeing to deploy them proves the market is catching up.

This is the moment when Serve transitions from a single-platform operator to a cross-network logistics AI company, supplying automation to the two largest delivery ecosystems in America.Image
$SERV $UBER $DASH is quietly becoming the universal AI layer for last-mile delivery

Uber and DoorDash are the two biggest delivery networks in the United States. Both are now using Serve Robotics’ autonomous delivery platform. This is the first time one robotics company has integrated at scale with both giants.

Serve’s model is simple but powerful. The same fleet of robots can dynamically switch between orders from Uber Eats and DoorDash, maximizing utilization and minimizing idle time. Every completed trip feeds new data into the autonomy stack, improving Serve’s AI foundation models and driving down cost per delivery.

Uber’s ownership stake ensures Serve is deeply embedded in the largest on-demand ecosystem in the world. DoorDash’s new partnership adds the second major leg of the market and signals that the company is ready to operate as a shared infrastructure layer rather than a captive service.

In effect, Serve is building what robotaxis could never deliver: a real, revenue-generating AI logistics network operating daily in U.S. cities. Two platforms, one fleet, one brain. That is how physical AI scales.Image
$SERV Serve isn’t just sidewalk robots anymore: drones, couriers, sanitation, and hospital AI robotics are next

Serve CEO Ali Kashani said, “To me, food delivery is the book to our Amazon. We’re just starting with that as a first market, but there is almost infinite demand for where we can go.”

The company’s roadmap extends far beyond restaurants. Serve plans to adapt its platform for groceries, prescriptions, and hospital logistics, moving critical items through campuses, malls, or factories using the same AI navigation stack that powers its delivery robots.

Kashani also confirmed that Serve is testing drone delivery integrations with partners such as Flytrex and Alphabet’s Wing, expanding into airborne autonomy for faster, lightweight dispatches. The same AI foundation models that teach a sidewalk robot to avoid pedestrians can teach a drone to navigate safely through complex urban airspace.

Serve is building a single autonomy layer that can move anything, anywhere. Sidewalks were only the training ground. The next phase is every surface and every altitude.Image
$SERV $GOOG Serve is taking autonomy airborne with its new drone partnership with Google

Serve Robotics is now integrating its sidewalk robots with Wing Aviation, an Alphabet company, to create a hybrid ground-to-air delivery network. The robots handle restaurant pickups, then hand off orders to drones for the final flight segment. This extends the average delivery radius from around 2 miles to as far as 6 miles, while cutting total delivery time and cost per order.

The pilot program in Dallas is already active. Serve’s robots retrieve food directly from restaurant partners, bring it to designated drone loading zones, and coordinate the handoff with Wing’s autonomous aircraft. The drones then complete the delivery in minutes, bypassing traffic and expanding the reachable market.

This partnership is not a small experiment. It represents a multi-modal autonomy system that connects sidewalk robots, road-based navigation, and aerial logistics under one unified AI brain. The same foundation models that train robots to detect obstacles on sidewalks are now being adapted to teach drones to plan safe, efficient aerial routes.

Serve’s long-term vision is to create a single, adaptable autonomy layer that can move goods by land or air. The company is proving that true “physical AI” is not limited to one form of mobility but can extend across every mode of delivery.Image
$SERV The Magna partnership proves $SERV is becoming the OS for AI robotics

Serve’s technology is now licensed and manufactured by Magna International, one of the largest automotive and robotics suppliers in the world. Magna builds vehicles and automation systems for Tesla, GM, and BMW, and is now producing Serve’s next-generation delivery robots while adopting its autonomy stack for new applications.

The partnership gives Magna a license to use Serve’s AI platform to develop entirely new robots and logistics systems. That means Serve’s navigation and perception engine can live inside robots far beyond food delivery, including industrial, commercial, and consumer uses.

CEO Ali Kashani said, “This collaboration positions our proprietary robotics technology as a platform upon which new robots can be built.” Magna Executive VP Matteo Del Sorbo added, “We are excited to continue collaborating with Serve, leveraging our manufacturing and technical expertise to help fuel Serve’s growth potential.”

Serve is no longer just a delivery company. It is becoming the operating system for physical AI, powering the next generation of intelligent robots that move goods, materials, and everything that makes cities function.Image
$SERV is led by one of the strongest robotics and AI teams in the industry

Serve Robotics is not just a tech story. It is a leadership story. The company is run by pioneers with decades of experience in robotics, autonomy, and scaling real-world systems.

Dr. Ali Kashani, CEO and co-founder, holds a PhD in robotics and computer vision. He led the Postmates X robotics division that built the original Serve robot and spun the company out of Uber in 2021 to keep it platform-agnostic. A TED speaker and EY Entrepreneur of the Year 2025, Kashani is known for his optimism about AI’s potential to improve quality of life. “Every major invention was feared at first. AI is going to improve healthcare, quality of life, everything,” he said.

Anthony Armenta, Chief Software and Data Officer, was previously CTO of GM’s BrightDrop, VP of Engineering at Postmates, and a senior leader at Anki Robotics. He brings 30 years of experience in autonomous systems and logistics technology. “Serve has built a leading autonomous delivery robot, demonstrating what’s possible for real-world performance,” he said.

Euan Abraham, VP of Hardware Engineering, came from GoPro and leads the design of Serve’s third-generation robots. He integrated Ouster LiDAR and sensor redundancy to achieve Level 4 autonomy. “We continue to be impressed by Ouster’s quality and reliability,” Abraham said.

The team expanded further with top engineers from Vayu Robotics, including former Velodyne CEO Dr. Anand Gopalan, a pioneer in perception hardware and autonomy. Legendary investor Vinod Khosla also joined Serve’s Advisory Board, bringing decades of deep-tech insight.

Serve’s board includes Sarfraz Maredia, Uber’s VP of Delivery, and investors such as Uber, NVIDIA, and 7-Eleven’s corporate fund. With more than 300 employees and a concentration of robotics PhDs and senior engineers, Serve has one of the highest “talent densities” in the sector.

This is a team that has already built and commercialized robots at scale. They are now building the future of physical AI.Image
$SERV has the balance sheet, growth, and platform to scale fast

Serve posted Q2 revenue of $642,000, up 46% quarter over quarter, with management guiding 170% to 215% year-over-year growth for Q3 2025. The company reiterated a $60–80 million annualized revenue target once its 2,000-robot fleet is fully utilized.

Serve ended Q2 with $183 million in cash and marketable securities, providing a clear runway through 2026. This liquidity allows it to self-finance robot production and expansion without relying on near-term dilution or debt.

Each robot generates recurring delivery income through partners like Uber Eats and DoorDash, while the platform adds high-margin software revenue through licensing agreements such as Magna International. As fleet utilization rises, unit economics improve dramatically.

It is still early days financially, and I would expect some capital raising, but the platform is remarkable. Serve has the foundation, partnerships, and scale potential to redefine last-mile logistics. For me, this earns a speculative buy rating within my AI portfolio.Image
$SERV Piling up results and achievements FAST! Image
My take: $SERV is the most promising small-cap AI robotics stock on the market

Serve is not just a delivery company. It is building the operating system for physical AI, a full-stack robotics platform that combines NVIDIA-powered autonomy, foundation model learning, and real-world data at scale. The same software that drives burrito deliveries today could one day power drones, hospital couriers, and factory logistics.

The company has tier-one validation from $UBER Uber, $DASH DoorDash, $NVDA NVIDIA, Magna, and legendary VC Vinod Khosla. It has achieved commercial Level 4 autonomy, strong early revenue growth, and a cash position that allows it to execute aggressively without dilution.

This is how big technological shifts begin: small teams, proven tech, and an expanding data flywheel. Serve’s robots are already navigating real cities and generating the datasets that will train the next generation of AI robotics.

It is still speculative, but in my view, $SERV represents the clearest early-stage exposure to the physical AI revolution. I am adding it to my Speculative AI Portfolio alongside $PATH, $RXRX, and $ASAN as a long-term asymmetric bet on the future of intelligent robotics.Image
$SERV I'm long Serve Robotics! Image

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with J Keynes

J Keynes Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @JKeynesAlpha

Sep 27
$RXRX 🚀Recursion Is The Next Moonshot In My Speculative AI Portfolio: Nvidia Is In, Big Pharma Is Paying, And Wall Street Has No Clue 🚀

This stock is sitting at all-time lows, but under the hood it’s building the most complete AI-first drug discovery platform in existence. The third pick in my speculative AI portfolio is Recursion Pharmaceuticals.

This is a stock that almost no one in the AI trade is watching right now, and that’s exactly the point. It sits near all-time lows, priced like another cash-burning biotech. Yet underneath, it’s building what I believe could be the most complete AI-first drug discovery platform in existence. NVIDIA is a holder, Big Pharma is already paying to use it, and the story is only beginning.

When I built this speculative AI portfolio basket, I wasn’t looking for polished blue chips. I wanted companies with asymmetric payoff structures: ones the market has ignored, where the downside is largely capped and the upside is multiples.

That’s why I started with $PATH, then $ASAN. Both had underappreciated AI leverage in enterprise workflows (you can learn about them in my feed).

For my third slot, I turned to a very different sector: medicine. I am already an investor in $TEM (Tempus AI) in my "AI Leaders Portfolio" which is building the clinical data infrastructure for AI in healthcare. And $RXRX complements my Tempus AI thesis perfectly and is already a large data customer of Tempus AI. Together they represent what I believe is the inevitable future: AI will transform medicine in every way, from diagnosis to drug discovery.

What makes RXRX unique is its AI-driven wetlab operating system. The company runs millions of automated experiments each week, generating one of the largest proprietary biological datasets in the world. These are not just simulations but live-cell assays, gene edits, and chemical perturbations captured through high-content imaging and structured into more than 60 petabytes of data. That information powers AI models that can surface new drug targets, design molecules to prosecute them, and guide clinical trial strategy. According to CEO Chris Gibson, the ultimate vision is to build a Virtual Cell, a computational model of biology so predictive that most of discovery can be done in silico before any patient is ever touched.

NVIDIA’s investment, combined with partnerships from Roche, Sanofi, Bayer, and Merck KGaA, shows that the most sophisticated players in tech and pharma see the potential here. RXRX has the capital and the collaborators to prove its approach at scale, and if it succeeds, this could be one of the most important companies in the future of medicine. (Tempus AI), which is building the clinical data rails for AI in healthcare. RXRX complements that thesis perfectly. Together they represent what I believe is the inevitable future: AI will revolutionize medicine in every way, from diagnosis to drug discovery.

Let's dive in 🧵Image
$RXRX The Vision: Solving "Eroom’s Law" with an AI Wetlab OS

Drug discovery is broken. Costs rise every year, timelines stretch longer, and the number of new medicines per dollar spent has fallen for decades. This slowdown is so well-known that it has its own name: Eroom’s Law, the inverse of Moore’s Law. As computing power has exploded, biology has resisted productivity gains. That is the problem Recursion set out to solve.

The answer, according to CEO Chris Gibson, is to industrialize biology the same way other industries embraced automation and simulation. That means moving away from the artisanal, trial-and-error methods that dominate pharma and replacing them with a system that can run experiments at scale, capture the data in structured form, and feed it into AI models that get smarter with every cycle. Recursion calls this system the Recursion OS.

What makes this vision different is the scale at which it operates. The company runs millions of live-cell experiments each week, generating more than 60 petabytes of proprietary data. Instead of starting with a hunch, the OS surfaces connections and pathways humans would miss. It can then loop directly into generative chemistry to design new molecules and into ClinTech to optimize trial strategies. The goal is not to make biology incremental. The goal is to make it programmable.

This is the foundation for RXRX’s potential 20x payoff. If it works, the company will not just produce a few drugs. It will have built the first AI wetlab operating system for medicine, with applications across every disease area. That is why NVIDIA invested, why Roche and Sanofi are paying to use it, and why RXRX at $5 is not the same story the ticker tells.Image
$RXRX Inside the AI Wetlab OS

Recursion’s edge starts in the wetlab. Most “AI biotechs” rely on public datasets or small outsourced experiments. RXRX industrialized the lab itself. According to the company, it runs more than two million automated experiments per week. These are not simulations. They are live-cell assays where genes are knocked out, compounds are added, and high-content imaging captures how cells respond. Every perturbation becomes structured data, adding to a dataset that now exceeds 60 petabytes.

The process is designed to be self-reinforcing. Recursion describes it as a cycle of wet-lab validation and dry-lab prediction. Robots handle cell culture, imaging, and perturbations with minimal human input. Each round of experiments feeds into AI models that generate new hypotheses, which then trigger the next set of automated experiments. The company’s stated goal is to model and simulate biology and chemistry at scale, so that discovery becomes programmable instead of artisanal.

One of the most powerful outputs of this system is the creation of “phenomaps.” By systematically knocking out every gene and mapping the resulting morphology, Recursion builds computational maps of biological function. These maps surface relationships no human could spot. For example, phenomaps revealed that RBM39 mirrored CDK12 activity without CDK13 toxicity, opening a path to a new oncology degrader program now in clinical trials. That discovery was not a guess. It was the product of industrialized wetlab automation feeding into AI analysis.

This is why RXRX stands apart. Algorithms alone are commoditized. The real moat is the ability to generate proprietary biological data at scale, week after week, in a way that compounds over time. That is the foundation for why RXRX has a chance to be a 20x play.Image
Read 17 tweets
Aug 31
$PATH 🚀 From bots to the Brain of Enterprise AI: UiPath is My First Speculative AI Portfolio Pick

Today, as promised I'm kicking off the Speculative AI portfolio with $PATH. You already know my AI Leaders portfolio, $TEM, $GTLB, $MDB, $ESTC, and $NBIS, the backbone of small and midcap AI infrastructure.

Tempus is healthcare data pipelines, GitLab is DevSecOps AI native, MongoDB and Elastic are the search and memory layer for Vector databases and bringing enterprise data to bespoke LLM and Agentic apps, and Nebius is contrarian, explosively growing, full stack infrastructure. Those are the core compounders.

But the Speculative AI portfolio is where I take shots on names that could re-rate big if their products cross the chasm. UiPath therefore gets the first slot in my list.

UiPath is trying to transform itself from an RPA vendor into the orchestration brain of enterprise agentic AI, and it seems more than a cosmetic rebrand.

In May 2025 on the Q1 FY26 earnings call, the company officially launched its Agentic Automation platform, which CEO Daniel Dines called “one of the most important and successful product introductions in our history.” He told investors, “together, we’ve turned the promise of Agentic automation into a powerful reality, and we are just getting started,” describing it as “how UiPath transforms work by unifying AI agents, robots and people into a single intelligent system.”

The Agentic Automation platform is not a single SKU but a suite of products that includes Agent Builder, Maestro, the AI Trust Layer, Autopilot, Agentic Testing, and the next-generation Intelligent Extraction Processing (IXP) solution. Dines laid out the “five powerful advantages” that give UiPath a right to win: “our extensive installed base of robots… our ability to bridge deterministic automation and probabilistic automation… our vendor agnostic architecture… our secure, enterprise grade governance… and our fully integrated end-to-end platform.”

By timing this launch with general availability of core products like Agent Builder and Maestro, and moving Maestro into preview in March 2025, UiPath is signaling that the transition is not theoretical. They are positioning themselves as more than a task automation vendor. They are claiming the role of enterprise control plane where AI agents are created, orchestrated, governed, and scaled into production🧵Image
$PATH The evidence of an AI turnaround is tangible. Customers have already created thousands of autonomous agents and generated over 250,000 agent runs with Agent Builder.

Maestro has powered more than 11,000 process instances.

A Fortune 15 healthcare company signed a multiyear, multimillion dollar deal to expand on UiPath’s agentic platform.

A global apparel retailer automated its entire support workflow using Maestro.

The U.S. Air Force launched its “Agentic Airmen Initiative” to embed agents into operations for 70,000 personnel.

Deloitte is co-selling UiPath’s agentic ERP model to a Fortune 20 oil and gas company preparing to automate 70 percent of manual testing in one of the world’s largest SAP projects.

The evidence is piling up.Image
$PATH The CFO’s Reinforcement

At the Mizuho Technology Conference on June 11, 2025, CFO Ashim Gupta echoed Dines and pushed back on AI hype directly: “Strong contacts and agreements with GSIs, good momentum with partners like Deloitte for Agentic ERP, really good momentum across the board, but it was underscored and highlighted by the launch of our Agentic platform. I know in the world of AI, a lot of people ask what’s real and what’s hype. We had the most significant product launch for us, tangible software that our customers had been previewing the previous three months, that is now in GA, that allows them to build and deploy agents at scale in conjunction with the rest of our automation platform.”

He went further: “We launched Agentic Orchestration, Agentic Testing, and Agentic Automation that allows humans, robots, and agents to be working together and really bringing that to our customers.” He emphasized scale, noting that UiPath now has “hundreds of proof of concepts” in progress and that the number one customer request is for more presales engineers, a sign customers want pilots and production runs, not marketing decks.

Gupta described UiPath’s role as “the Switzerland for Agentic,” explaining that while some companies will build agents for narrow verticals like CRM or HR, UiPath’s value is to orchestrate across the 10,000–15,000 applications most enterprises already run. Governance is central: “If you’re in health care doing claims processing, governance and controls are a significant part of your decision making, and that’s where we are differentiated.”Image
Read 14 tweets
Jul 14
$TSLA $TEM $NBIS $CRWV $NVDA $GOOG $QS $JOBY $OKLO AI, Robots, Healthcare, Batteries, and Transportation: Khosla's Future Investing Playbook to Torch the Fortune 500 🤖💸🔋⚕️🚗☢️

Just finished the Vinod-Khosla-with-Jack-Altman interview and my head's buzzing. Khosla opens by saying, "Almost every job is being reinvented… any economically valuable job humans can do, AI will be able to do eighty percent of it within the next five years." Pause and sit with that. Five years. This is the guy who funded Sun, Juniper, and OpenAI back when each looked absurd.

What's hitting me is that Keynes called this exact moment in 1930. He predicted humanity would solve its "economic problem" within 100 years. Khosla is saying we'll do it in 5. That's 2030: right on time.

Keynes anticipated exactly this moment: "We are being afflicted with a new disease... namely, technological unemployment. This means unemployment due to our discovery of means of economising the use of labour outrunning the pace at which we can find new uses for labour." He called it a "temporary phase of maladjustment" before humanity solves "its economic problem."

If Khosla's clock is right, we are living through the single fastest repricing of labor since the steam engine. But this isn't just about automation; it's about reaching what Keynes called the end of the "permanent problem of the human race." The economic problem that has driven all of human evolution may be solved within the decade.

But here's where it gets unsettling. Keynes worried about what happens when economic necessity disappears; would we face a collective "nervous breakdown"? He observed the wealthy classes of his time, calling them "our advance guard... spying out the promised land," and found their example "very depressing." They had failed to solve the fundamental question: how do you live meaningfully when survival is guaranteed?

The investment implications are staggering. If Khosla's timeline holds, we're not just repricing labor; we're repricing the entire social contract. This is where radical technological innovation will change the very fabric of capitalism itself. Investors will want to profit from this transformation, and the communists have it all wrong: capitalism IS the revolution. The market mechanism is what's driving us toward post-scarcity, not fighting it.

Keynes distinguished between "absolute needs" (food, shelter, basic comfort) and "relative needs" (status, superiority over others). The absolute needs market gets commoditized to near zero; the relative needs market becomes everything.

This suggests a brutal K-shaped economy: businesses serving absolute needs face margin compression toward zero, while anything touching relative needs (luxury, status, uniqueness, human connection, meaning) captures exponential value. But to get to that zero marginal cost production economy, we have to invest in the companies building every sort of relevant future technology. These are the explosive places that create the post-scarcity infrastructure first.

Every thesis about growth stocks, every DCF we build on ten-year head-count curves, needs a rewrite. Your models need a philosophical upgrade: What does this company sell when labor costs nothing? And what companies will sell the core technologies, the picks and shovels, and the daring innovations to fill in the core infrastructure of the world to come?

So, what companies position us for this new economy that hasn't been built yet? Even if Khosla's 5-year timeline is wildly optimistic, and oh boy, it's coming regardless. 🧵Image
$TSLA $TEM $NBIS $CRWV $NVDA $GOOG Khosla doubles down on the social fallout. He predicts that somewhere around 2030 the sheer productivity gain flips the labor equation: "The need to work will go away. People will work on things because they want to, not because they need to pay the mortgage."

This is exactly what Keynes envisioned in 1930. He predicted we'd reach a fifteen-hour work week where people would "do more things for ourselves than is usual with the rich today, only too glad to have small duties and tasks and routines." Work becomes voluntary self-expression rather than economic necessity.

But Keynes saw the economic disruption clearly. When work becomes voluntary rather than necessary, the entire labor market structure collapses. This echoes Karl Polanyi's insight from "The Great Transformation": labor was never meant to be a commodity. Polanyi argued that treating human work as a market good was always an artificial construct that would eventually break down.

What we're witnessing with AI is the final stage of that breakdown. When machines can perform 80% of economically valuable tasks, the fiction that human labor is a scarce commodity ends. This isn't just technological unemployment; it's the obsolescence of employment as an organizing principle for society.

The investment implication is massive: we're moving from a labor-scarce to a labor-abundant world. Companies built on exploiting labor arbitrage become worthless. Companies that can deploy capital and technology without scaling human workers become exponentially valuable. The entire framework of "revenue per employee" becomes meaningless when the denominator approaches zero.

I'm not here to debate universal basic income. What matters for investors is that pricing power for routine expertise collapses. If the marginal cost of a "junior analyst" or a "primary-care triage" visit goes to near zero, firms that sell packaged know-how must either pivot to data moats or watch revenue melt. Anyone holding legacy service businesses should reread that sentence until it hurts.

This creates a brutal bifurcation in AI investments. You want to own the companies building the AI tools that destroy pricing power, not the companies getting destroyed by them. McKinsey's consulting model dies; Palantir's $PLTR data integration platform thrives. H&R Block's tax prep gets commoditized; Intuit's $INTU financial platform captures the value. Traditional medical groups get squeezed; $TEM Tempus AI's diagnostic tools extract the margin. The playbook is clear: own NVIDIA's $NVDA infrastructure, Oracle's AI Stargate project, OpenAI's models, Anthropic's safety research, xAI's compute power, Palantir's government contracts, $BBAI AI defense applications (should they scale). These companies scale infinitely without human bottlenecks.

Stargate itself exemplifies this perfectly: $ORCL Oracle's $500 billion infrastructure joint venture with OpenAI isn't just building data centers, it's constructing the physical foundation for Keynes' post-scarcity economy. When Oracle, and neocloud AI stack companies like $NBIS and $CRWV, can rent computing power and AI deployment expertise that replaces entire workforces, the company collecting that rent captures exponential value while labor costs approach zero.Image
$TSLA $TEM $NBIS $CRWV $NVDA $GOOG
The conversation shifts from knowledge to robotics.

Khosla's robotics call is the spiciest I've heard this year:
"We'll have the ChatGPT moment for general-purpose humanoids in the next two to three years… almost everybody in the 2030s will have one at home."

Two to three years for the tipping point, not ten. He is talking about robots that learn, not bolt-tightening arms. The gating item is intelligence, not hardware; the learning stack just has to click once, then scale like software. If that lands, everything that still relies on human physical throughput—manufacturing, logistics, even household chores—re-prices overnight. Battery density, drivetrain motors, edge AI inference chiplets $NVDA $AMD $AVGO $NVTS: those become the new oil.

The battery piece is central. Solid state batteries like those $QS QuantumScape is developing solve the energy density problem that's held robotics back for decades. These aren't just improvements; they're order of magnitude leaps in power to weight ratios that make truly mobile, autonomous robots economically viable. Tesla will deploy robot workers at scale, and a new round of SPACs and startups will follow around AI robotics of every kind. But investing in the battery infrastructure is just as crucial as betting on the robotics companies themselves. The suppliers that make solid state batteries work (the materials, the manufacturing equipment, the charging systems) will capture massive value as every physical task gets automated. You can bet on it! QuantumScape is here and is about to break out, but other names in the robotics space will emerge and will need to be found, but the thesis is clear: solid state batteries are the foundation that makes Khosla's robot future possible.

Khosla Ventures was an early investor in QuantumScape, betting on their solid-state battery breakthrough before most investors understood the transformative potential for both electric vehicles and the robotics revolution ahead.Image
Read 11 tweets
Jun 29
$TEM 🚨 TEMPUS AI: THE NEXT TRILLION DOLLAR COMPANY? 🚨

Just listened to Tempus CEO Eric Lefkofsky drop absolute BOMBS on this Scottish Mortgage podcast (less than 2700 views on YouTube as of today!). He literally said Tempus AI has "trillion dollars a year of cash flow" potential and compared Tempus to Apple, Google, Amazon & Nvidia. I overlooked this podcast from late February before but it gives insanely valuable color into Lefkofsky's vision and the AI developments coming down the line at Tempus. Sometimes CEOs hype, and Lefkofsky admits it might not happen, but he gives good reason to think that Tempus could be the next megacap company we all hope it will be.

After hearing this, I'm more than ever convinced Tempus is a solid 10x + investment for the future. Scottish Mortgage was an early 2018 investor and Lefkofsky breaks down everything from their $1.2B AFIB algorithm opportunity to his "Nvidia moment" prediction to how they plan to extend human lifespan by A DECADE while saving $3 trillion in healthcare costs.

This thread will blow your mind. 👇Image
$TEM THE MASSIVE OPPORTUNITY: The Scale Is Inhuman

Lefkofsky just dropped this bomb about Tempus' potential. He said "I suspect one day there will be a Tempus-like company that is every bit as big as Apple or Google or Amazon or Nvidia or whatever. Because it will be routing tens of millions of patients around the world to the right therapy and getting paid a small toll every time it does that."

He's literally comparing Tempus to the biggest tech companies in the world. The audacity is incredible but when you hear his reasoning, it starts to make sense. Healthcare is the biggest industry in the world, and if this technology is going to be impactful in the world then it has to be impactful in healthcare. It's not surprising Google remains a shareholder and early seeder of Tempus.Image
Image
$TEM From Personal Tragedy to Global Mission

The origin story is deeply personal and explains Lefkofsky's relentless drive. About ten years ago his wife was diagnosed with breast cancer. As a tech entrepreneur since 1999, he was amazed at how little data and technology were part of her care.

He said: "I was kind of amazed at how little data and technology were part of her care and a part of really anyone that was being treated with cancer. And so I started thinking, this is crazy that we're giving truck drivers that are delivering pallets of water bottles more technology than we're giving physicians who are making life and death decisions."

The genomic sequencing reports coming back had no idea who his wife was, didn't know if she was male or female or old or young, didn't know what drugs she had taken. They were recommending drugs she had already failed on and clinical trials she wasn't eligible for. This personal frustration with the healthcare system's data silos became the founding mission of Tempus.Image
Read 17 tweets
Jun 26
$QS Solid-State Is No Longer a Myth: QuantumScape Might Be the Trade of the Decade

This isn’t another pipe dream about solid state batteries arriving “someday.” QuantumScape CTO Tim Holme laid out their case on a recent "Batteries Included" podcast, and if you understand what was said, you’ll realize they may be sitting on the most important battery tech in a generation. This isn’t just a science project anymore. This is real data, real partnerships, and a roadmap that leads directly to mass production. If you care about energy density, fast charging, cost, longevity, or simply being early to a generational leap, keep reading.Image
$QS A Battery That Changes All the Rules

Dr. Tim Holme, co-founder and CTO of QuantumScape, opened the interview with a statement that sets the tone for everything that follows. “A solid state battery could potentially be safer, but even more importantly could potentially enable a lithium metal anode. A lithium metal anode in a battery would give the battery higher energy density and potentially also faster charge, better safety, better lifetime. It’s rare that a battery can advance in all of the metrics that are important to battery applications simultaneously. The last time that really happened was when the lithium ion battery was introduced in 1991.”

What he’s describing isn’t incremental. This is a platform shift in battery chemistry with the power to improve every core metric at once. That almost never happens.Image
$QS World-Beating Energy and Power Numbers

QuantumScape’s A and B sample results are not just solid. They crush the current generation. Holme gave the full numbers: 844 watt-hours per liter and 301 watt-hours per kilogram. For comparison, today’s top energy-dense lithium ion cells hit the mid 200s for Wh/kg and around 720 to 740 Wh/L. Holme explained, “Our cell has the power that exceeds the power from a power cell… We’re beating the energy cells on energy and also beating the power cells on power at the same time.”

Most battery chemistries force tradeoffs. You can have energy or power, but not both. QuantumScape claims it delivers both in one cell, with data to back it up.
Read 15 tweets
Sep 28, 2024
$ionq 🚀 IonQ is taking quantum computing to new heights with groundbreaking deals, including a $54.5M contract with the U.S. Air Force Research Lab and other federal partnerships, such as the $40M ARLIS Quantum Computing Intelligence contract for the DoD, positioning itself as the go-to for quantum networking and quantum computing power for national defense. 🌍💻

With the upcoming AQ 64 Tempo Enterprise systems and a $95M revenue forecast, IonQ is rapidly expanding its reach. But what’s behind this surge? Let’s dive into the political support that’s fueling IonQ’s success, from key lawmakers backing quantum innovation to growing federal contracts.

👀 The thread ahead will peel back the curtain on the politicians driving federal interest in IonQ’s technology. From quantum cybersecurity to cutting-edge defense applications, bipartisan support is making IonQ the face of America’s quantum future.

A deep dive 🧵

#IonQPolitics #QuantumInnovation #TechPolicy #FederalContracts #QuantumNetworking

#QuantumComputing #IonQ #QuantumDefense #TechPolicy #AQ64
$ionq Ben Cardin, D-Maryland Image
$ionq Cardin has been a key advocate for Maryland’s leadership in quantum technology, playing a pivotal role in securing federal support for projects like @IonQ_Inc. His recent announcement of $3.5M in funding for the UMD/NIST Joint Quantum Institute and QuICS demonstrates his commitment to advancing quantum computing research. Cardin’s backing of @IonQ_Inc and its partnerships with UMD, including the $20M National Quantum Lab (QLab), highlights his dedication to positioning Maryland as a global hub for quantum innovation. This federal support not only strengthens academic research but also ensures IonQ’s quantum technologies continue to thrive and impact critical sectors like AI and national security. #QuantumComputing #MDLeadershipImage
Read 18 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(