Watchlist

Watchlist

Companies we are researching but don't own. For each one: the thesis, the bull and bear cases, upcoming catalysts and the open questions, with every claim linked to its source.

Names we are researching, not positions. Tickers only. Every claim links to its source; quotes are copied word for word from the linked page. Last updated 2026-10-23. Prices are live quotes and may be delayed.

T

AT&T

NYSE
–Live price loads in your browser

ThesisJensen Huang argues that telecom operators will turn their land, power and connectivity into distributed AI infrastructure. If that happens, a carrier with a large fiber and wireless footprint could carry and run AI workloads, not just traffic. AT&T is the U.S. carrier we want to test that idea on.

“Telecom networks were built to carry the world’s conversations. Now, they will also carry the world’s intelligence. Leading global operators are turning their land, power and connectivity into a distributed AI infrastructure — bringing intelligence closer to the industries, people and robots that use it. With Nokia, we’re extending accelerated computing into the radio network itself, creating the foundation for AI-native 5G.”

Jensen Huang, NVIDIA founder and CEO · 21 Oct 2026, GTC Berlin (NVIDIA press release)1Verified word for word on NVIDIA's own release.

“Every single telco in the future, and I predict this, every single telco in the future, with 100,000 sites representing some 100 gigawatts of distributed telco base stations and data centers, will in the future evolve into a distributed AI network.”

Jensen Huang, GTC Berlin keynote · 21 Oct 2026, as quoted by Yahoo Finance2Secondary: quoted by Yahoo Finance from the keynote. We have not yet checked it against the keynote video.

Wording check “AI factories” is not Jensen's wording about telcos in the sources we checked. His own words are “distributed AI infrastructure” (NVIDIA release) and “distributed AI network” (keynote, as reported). The phrase “telco-led AI factories” comes from an NVIDIA blog post by Chris Penrose, not from Jensen: “Eighteen telco-led AI factories powered by NVIDIA now span five continents.” (11 Jun 2025).3

AT&T's actual NVIDIA ties

  • 17 Mar 2026: AT&T and Cisco announced a network-driven edge AI service for enterprise IoT. It runs Cisco AI Grid with NVIDIA on AT&T's dedicated IoT core, powered by NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.4
  • Live use: a public-safety demo with Linker Vision at AT&T Discovery District in Dallas, plus a pilot with TanMar Companies in Eunice, Louisiana.4
  • NVIDIA's blog on telecom “AI grids” covers the same AT&T, Cisco and NVIDIA work, with a quote from AT&T's Shawn Hakl.5
  • Connected AI for Manufacturing: AT&T with MicroAI, NVIDIA and Microsoft.6

Bull case

  • AT&T already sells an NVIDIA-powered edge AI service with Cisco, so it is not starting from zero.4
  • The asset base Jensen describes: AT&T reached more than 39 million fiber locations in Q3 2026 and targets more than 60 million by end of 2030. CEO John Stankey: “Our asset base positions us well for the AI era, with a leading position in fiber – the best technology available.”7
  • Q3 2026: revenue $32.3 billion (+5.3%) and adjusted EPS of $0.65.7

Bear case

  • AT&T is not among the operators named in NVIDIA's 21 Oct 2026 telecom announcement (it names T-Mobile, BT, Deutsche Telekom, Elisa, Orange and others).1
  • NVIDIA's announced AI-RAN deployments use Nokia's anyRAN software (“nearly 20 telecom operators” globally). AT&T picked Ericsson for its Open RAN rollout in Dec 2023, so the most visible AI-RAN path does not run through AT&T's main radio vendor.18
  • Balance sheet: total debt $159.6 billion and net debt $149.8 billion at 30 Sep 2026, and AT&T is lowering long-term capital investment to $22–23 billion a year in 2027–2028.7

Key catalysts

  • 21 Oct 2026 (passed)Q3 2026 results reported.7
  • Date not announcedQ4 2026 results. AT&T has not announced the date yet.
  • PendingSale of the Mexico wireless business (agreement signed 13 Oct 2026).7
  • 2027Industry test: BT's live AI-RAN trials in Birmingham and Deutsche Telekom's field trials. Watch whether a U.S. carrier besides T-Mobile joins.1

Research to do

  • Does AT&T plan AI-RAN with Ericsson and NVIDIA, or anything beyond the Cisco IoT AI Grid?
  • How much revenue does the AI Grid / edge AI service bring in? It isn't disclosed yet.
  • Check Jensen's keynote line against the GTC Berlin video, word for word.
  • Did AT&T appear on any GTC keynote slide? Not checked: we used press releases, not the slides.
  • Does AT&T rent out sites, power or fiber to AI data centers, and on what terms?

Research areaTelecoms NVIDIA has named as AI-RAN or AI-factory partners

From NVIDIA's own GTC releases and blog posts, plus each operator's announcement. “Verified” means we read the primary source linked. We did not review the keynote slides, so “shown on screen” is not checked for any name.

CompanyTickerWhat was announcedDateStatus
T-Mobile USTMUSWorld's first AI-RAN call on a commercial network (Bellevue, Washington) with NVIDIA and Nokia. AI-RAN trials with Nokia and NVIDIA were first announced at GTC Washington, D.C.1921 Oct 2026 (call); 28 Oct 2025 (trials announced)Verified
Nokia (equipment maker, not an operator)NOKNVIDIA invests $1 billion at $6.01 per share; Nokia builds AI-RAN products on NVIDIA's Aerial RAN Computer.928 Oct 2025Verified
Deutsche TelekomDTE (Xetra)Expanding its NVIDIA-powered Industrial AI Cloud in Munich; AI-RAN validated in labs with field trials planned for 2027.11021 Oct 2026Verified
OrangeORA (Euronext Paris)Testing Nokia 5G software on NVIDIA ARC; hosting Nemotron models on its Live Intelligence platform. Joined the NVIDIA Cloud Partner program at GTC Paris.1321 Oct 2026; 11 Jun 2025Verified
BT GroupBT.A (London)Live AI-RAN trials in Birmingham with Nokia and NVIDIA planned for 2027; building Nemotron agents for field work.121 Oct 2026Verified
ElisaELISA (Helsinki)Europe's first AI-RAN field trial on NVIDIA ARC.121 Oct 2026Verified
SwisscomSCMN (SIX)Enterprise AI services (GenAI Studio) hosted on its sovereign AI factory built on NVIDIA DGX SuperPOD.311 Jun 2025Verified
Fastweb + Vodafone (Italy)Owned by Swisscom (SCMN); no own listingBuilt its MIIA Italian language model on an NVIDIA DGX supercomputer; named among the AI infrastructure expansions at GTC Berlin.311111 Jun 2025; 21 Oct 2026Verified
TelenorTEL (Oslo)Expanding Norway's sovereign AI infrastructure, with a new AI data center running on renewable energy.311 Jun 2025Verified
TelefónicaTEFPiloting distributed edge AI across Spain with hundreds of NVIDIA GPUs.311 Jun 2025Verified
TIM (Telecom Italia)TIT (Milan)New NVIDIA-powered AI infrastructure.121 Oct 2026Verified
SoftBank Corp.9434 (Tokyo)AITRAS, an AI-RAN system running AI and 5G on NVIDIA GH200. Jensen: it “allows every telco to become an AI service provider”.12Announced by SoftBank, not at a GTC; we did not confirm which NVIDIA event it was tied to.13 Nov 2024Verified
Indosat Ooredoo HutchisonISAT (IDX)Zankore by Indosat: target of 1 GW of NVIDIA DSX AI Factory capacity, about 200 MW on GB300 NVL72 in H1 2027.13Announced through Nokia's newsroom, not at a GTC.6 Aug 2026Verified
SingtelZ74 (SGX)Joined the NVIDIA Partner Network Cloud Programme to bring AI to Singapore and Southeast Asia.14A Singtel announcement, not a GTC one.1 Feb 2024Verified
VerizonVZPrivate 5G and Mobile Edge Compute with NVIDIA AI Enterprise and NIM for on-site enterprise AI.15An edge AI product, not AI-RAN or an AI factory. We found no NVIDIA GTC release naming Verizon as an AI-RAN or AI-factory partner.17 Dec 2024Partly verified

Also named by NVIDIA on 21 Oct 2026 as building or expanding NVIDIA-powered infrastructure: CRA (Czech Republic), etisalat (UAE), Silknet (Georgia), Cassava (Africa), Kazakhtelecom and Orion Telecom (Serbia).1

Sources (15)
  1. NVIDIA newsroom: NVIDIA, Nokia and Leading Telecom Operators Turn Networks Into Distributed Intelligence Platforms (21 Oct 2026, GTC Berlin)
  2. Yahoo Finance: Nvidia's Jensen Huang Sees 100GW AI Opportunity In Telecom Networks (21 Oct 2026)
  3. NVIDIA blog: Leading European Telcos Build AI Infrastructure With NVIDIA (11 Jun 2025, GTC Paris)
  4. AT&T newsroom: AT&T Leads Industry Collaboration with Cisco and NVIDIA (17 Mar 2026)
  5. NVIDIA blog: NVIDIA, Telecom Leaders Build AI Grids to Optimize Inference on Distributed Networks
  6. AT&T blog: How AT&T is Leading the Next Era of IoT (2026)
  7. AT&T newsroom: AT&T Delivers Strong Third-Quarter 2026 Results (21 Oct 2026)
  8. AT&T newsroom: AT&T Collaborates with Ericsson to Accelerate Open RAN in the U.S. (4 Dec 2023)
  9. NVIDIA newsroom: NVIDIA and Nokia to Pioneer the AI Platform for 6G (28 Oct 2025, GTC Washington, D.C.)
  10. TelecomTV: Nvidia boasts greater AI-RAN and AI factory traction with telcos (21 Oct 2026)
  11. Swisscom: Swisscom completes acquisition of Vodafone Italia (2 Jan 2025)
  12. SoftBank Corp.: Announces Development of AITRAS, a Converged AI-RAN Solution (13 Nov 2024)
  13. Nokia newsroom: Indosat, Ooredoo Group, Nokia and NVIDIA launch Zankore by Indosat (6 Aug 2026)
  14. Singtel RE:AI newsroom: Singtel collaborates with NVIDIA to bring AI to Singapore and Southeast Asia (1 Feb 2024)
  15. Verizon newsroom: Verizon collaborates with NVIDIA to power AI workloads on 5G private networks (17 Dec 2024)
UBER

Uber Technologies

NYSE
–Live price loads in your browser

ThesisUber wants to be the network that robotaxi fleets plug into, with NVIDIA supplying the car computer and driving software. The open question is whether Uber keeps the rider relationship and the pricing power when the strongest robotaxi players own their vehicles, their software and their own apps.

Uber and NVIDIA: what was announced

  • 28 Oct 2025 (GTC Washington, D.C.): Uber to begin scaling a global autonomous fleet in 2027, targeting 100,000 vehicles over time, on NVIDIA DRIVE AGX Hyperion 10 with two DRIVE AGX Thor computers on NVIDIA Blackwell. The two companies are also building an AI data factory on NVIDIA Cosmos.1
  • Automaker and AV partners named in that release: Stellantis (with Foxconn), Lucid and Mercedes-Benz for passenger cars; Avride, May Mobility, Momenta, Nuro, Pony.ai, Wayve and WeRide as shared partners.1
  • 16 Mar 2026: NVIDIA's own L4 software (Alpamayo on DRIVE Hyperion) to run robotaxis on Uber, starting in Los Angeles and San Francisco in the first half of 2027 and scaling to 28 cities by 2028.2
  • 21 Oct 2026 (GTC Berlin): NVIDIA restated that Uber and NVIDIA are building a Cosmos-powered data factory, part of a partnership covering Hyperion-based vehicles and Uber's network.3

Bull case

  • One platform, many suppliers: NVIDIA lined up carmakers (Stellantis, Lucid, Mercedes-Benz) and AV companies on the same Hyperion design, which could give Uber many fleets to choose from rather than one.1
  • Concrete timeline: Los Angeles and San Francisco in H1 2027, then 28 cities by 2028, in phases.2
  • Zoox, which owns its own vehicles, still chose to put robotaxis on the Uber app: Las Vegas first, Los Angeles by mid-2027.4

Bear case

  • Uber's own 10-K names Waymo, Tesla and Zoox among current or future competitors, and notes: “Waymo has introduced a commercialized ridehailing fleet of autonomous vehicles on its own platform in addition to a fleet … through our platform”.5
  • Waymo plans to launch its own app in Austin and Atlanta in January 2028, ending exclusivity in the two cities where it is only on Uber. Uber shares fell more than 4% that day. Tesla, Zoox and others already run standalone apps.6
  • Price: in Obi's San Francisco data (1 Mar–14 Apr 2026), Tesla Robotaxi was still the cheapest option, roughly half the price of Waymo on average, and undercut Uber and Lyft per mile even after a 41% price rise.7
  • NVIDIA's stack is not exclusive to Uber: at GTC Berlin, NVIDIA said Bolt picked Alpamayo for a European robotaxi rollout of at least 25,000 Lucid vehicles built on Hyperion.3

Key catalysts

  • 4 Nov 2026, 8:00 a.m. ET (14:00 Madrid)Q3 2026 results conference call.8
  • Announced for summer 2026; not confirmed liveZoox robotaxis on the Uber app in Las Vegas.49
  • H1 2027First NVIDIA-software robotaxis on Uber in Los Angeles and San Francisco.2
  • Mid-2027Zoox on Uber in Los Angeles.4
  • January 2028Waymo's own app launches in Austin and Atlanta; the Uber contract runs through at least May 2028.6

Research to do

  • Has the Zoox ride option actually gone live in the Uber app in Las Vegas? The newest source we found (TechCrunch, 3 Sep 2026) still describes it as planned.
  • What does Uber earn per robotaxi trip compared with a human-driven trip?
  • Which automakers will supply the 2027 Los Angeles and San Francisco fleets, and who owns and finances the cars?
  • Robotaxi price data for Las Vegas, where Zoox, Waymo, Tesla and Uber all hold permits.
  • Watch the Q3 call (4 Nov) for fleet numbers and AV trip share.

Recent news

  • 22 Oct 2026Uber is rolling out ChatGPT Work company-wide, to more than 30,000 employees, after a pilot with 1,000. GPT-6.1 Sol is the default model; engineers keep using Codex.10

Research areaWhat Amazon is doing with Zoox

Where it runsLas Vegas: open to the public in the Zoox app. San Francisco: early riders let in from a waitlist (Zoox Explorers), with the service area quadrupled this spring. Austin and Miami: purpose-built robotaxis on public roads, first for employees and their families and friends. Testing also announced in San Diego and Houston.11129
Launch datesPaid rides in Las Vegas started 10 Aug 2026, after a federal exemption. It allows up to 2,500 vehicles a year for two years. Rides to and from Las Vegas airport (LAS) started 3 Sep 2026.139
Scale so farNearly 1 million riders across Las Vegas, San Francisco, Austin and Miami since the free service launched (as of Aug 2026).13
Vehicle productionHayward, California plant, opened Jun 2025, with capacity for more than 10,000 robotaxis a year at full scale. Zoox “owns, operates, and assembles” its robotaxis.14
Which appIts own Zoox app today. Uber deal (11 Mar 2026): Zoox on the Uber app in Las Vegas “starting this summer” and Los Angeles by mid-2027, while the Zoox app stays. As of 3 Sep 2026, TechCrunch still described Uber's Las Vegas robotaxi plans, Zoox included, as upcoming.49
PricingZoox said it aims to be competitive with “comfort” level ride-hailing prices. It did not give a base fare.13
ChipsMultiple NVIDIA GPUs in each vehicle, plus NVIDIA GPUs for simulation. Zoox CTO Jesse Levinson: “We’ve been using NVIDIA hardware since the very start”. Source dated Jul 2024; we did not find a newer statement on the current vehicle's chips.15
Sources (15)
  1. NVIDIA newsroom: NVIDIA Makes the World Robotaxi-Ready With Uber Partnership (28 Oct 2025, GTC Washington, D.C.)
  2. Uber investor relations: NVIDIA to Launch L4 Software-Driven Robotaxis on Uber Across 28 Cities by 2028 (16 Mar 2026)
  3. NVIDIA newsroom: European Mobility Leaders Build Safe Vehicles and Robotaxi Fleets on NVIDIA Hyperion (21 Oct 2026, GTC Berlin)
  4. Uber investor relations: Zoox and Uber Announce Strategic Partnership (11 Mar 2026)
  5. Uber Form 10-K for fiscal 2025 (SEC EDGAR), competition and risk factors
  6. CNBC: Uber and Waymo to end exclusivity arrangement in Atlanta and Austin (24 Jul 2026)
  7. Obi: Tesla Robotaxi Prices Are Up 41% In San Francisco (21 Apr 2026; data 1 Mar–14 Apr 2026)
  8. Uber investor relations: Uber Announces Date of Third Quarter 2026 Results Conference Call (5 Oct 2026)
  9. TechCrunch: Amazon's Zoox expands its robotaxi service to Las Vegas airport (3 Sep 2026)
  10. OpenAI: Uber brings ChatGPT Work to all employees with OpenAI (22 Oct 2026)
  11. Zoox: Las Vegas service page
  12. Zoox journal: Zoox expands robotaxi service and unveils new product features (23 Mar 2026)
  13. CNBC: Amazon's Zoox to launch paid robotaxi service in Las Vegas on Aug. 10 (5 Aug 2026)
  14. Zoox journal: Inside the Zoox robotaxi serial production facility (18 Jun 2025)
  15. NVIDIA blog: NVIDIA and Zoox Pave the Way for Autonomous Ride-Hailing (31 Jul 2024)

Signals

SIGNAL

OpenAI: internal models, self-improvement and chips

21 Oct 2026 · Sam Altman, in a 45-minute interview with Matthew Berman (as reported by Dealroom)1
  • Altman said OpenAI's real advantage is the models it runs internally, which he described as well ahead of what the public can access (Dealroom's paraphrase, not a direct quote).
  • On recursive self-improvement, his words were: “a very dangerous thing to push on”. He said it will need “speed limits” on how fast it is allowed to go.
  • Asked whether the lab that pushed hardest on self-improvement could pull permanently ahead, he said yes, “if people really pushed on RSI.”
  • He named OpenAI's own chips as its most important bet (Dealroom's paraphrase).

Quotes match Dealroom's article word for word. We have not watched the original interview video.

Sources (1)
  1. Dealroom News: Altman: OpenAI's edge is internal models, and self-improving AI needs "speed limits" (interview with Matthew Berman published 21 Oct 2026)

A watchlist is not a recommendation and we hold no position in these names unless they appear on What We Own. Not investment advice. Disclaimer.