Between 2017 and 2022, dozens of crypto projects promised to rebuild travel. Book a flight with Bitcoin. Own your hotel room as an NFT. Cut Booking.com out of the loop.

The companion card, The On-Chain Travel Stack, maps the names still running today. Strip it back to the projects that actually launched in this 2017-2022 wave and are still operating, and you are left with barely a quarter of the field. This is the other three quarters.

Here is what happened to the ones that didn't make it.

1. Winding Tree: the one that got closest

Raised $14.2M. Died July 2024. Seven years old.

Winding Tree was not vaporware. It ran live pilots with Lufthansa, Air France-KLM, Air Canada, Air New Zealand, Nordic Choice Hotels and citizenM. Hahn Air issued real airline tickets over its platform. The product worked.

It still lost.

CEO Maksim Izmaylov published a post-mortem in July 2024 that is the most honest document in this entire sector. His three reasons:

1. Crypto's sole real use case turned out to be gambling. The industry evolved in a direction that had nothing to do with travel distribution.

2. The travel industry used "blockchain" as marketing language while having zero interest in actual decentralization.

3. Even at major hotel chains, there was nobody on staff who could process a crypto transaction.

Their EY partnership reached a working, revenue-generating product. Then EY's management changed and it was cancelled.

One detail worth keeping: Winding Tree capped its own ICO. They raised $14.2M from about 7,000 backers, kept $10M, and routed the excess into a smart contract that bought back and burned tokens. In 2018. Almost nobody did this.

LIF today: $64K market cap, one exchange, $0 in daily volume. Their GitHub org contains a repository named "winding-down." The @windingtree handle now belongs to a stranger who registered it in May 2025.

2. Pinktada: resold hotel nights as NFTs

Raised ~$4.25M. Went quiet July 2023.

Genuinely clever mechanic: every non-refundable hotel night became a tradeable NFT. Cannot make the trip? Resell the night instead of eating the cancellation. Sell it at a profit and the hotel took 30% of your upside.

The Wall Street Journal covered it. It required two things that never showed up together: hotels willing to trade refund revenue for guaranteed sales, and a liquid market of buyers for one specific room on one specific night.

Ran out of runway as the NFT market collapsed and "NFT" became a liability word in consumer travel.

No shutdown announcement was ever made. Today pinktada.com redirects to an Indonesian online-slots affiliate page. PitchBook and Tracxn still list it as an active five-person company.

3. Aeron: pilot logbooks on chain

Raised $5.7M against a $30M cap. Last post April 2022.

The pitch was serious: falsified flight hours and maintenance records cause crashes, so put them on a blockchain. They shipped a working logbook app in 2018.

Aviation records are among the most heavily regulated data on earth. No regulator or airline was moving them onto a public chain. Zero adoption.

So they pivoted. Into a loyalty token. Then staking. Then a trading game. Binance delisted ARN in July 2020.

Their final public post, April 2022, opens by explaining the development team lost several key developers in Ukraine. The war wrote the last sentence, not the market.

ARN peaked at $8.67 in January 2018. Last recorded price: $0.0028. The @aeronaero account has been renamed to the display name "inactive" with a bio of four dashes.

4. GOeureka: zero-commission hotels, twice

Two token sales. Second life lasted 31 days.

The 2018 pitch was reasonable: let hotels list without paying Booking.com commissions, charge travellers 5% and waive it if they pay in GOT.

The 2018 ICO did not produce a working business. The team admitted it in writing: "there haven't been many updates from us over the past 12 months." So in July 2021 they migrated to Tezos, blamed Ethereum gas fees, and ran a second sale.

You can date the death from on-chain data. Their Tezos sale contract shows first activity 14 July 2021 and last activity 14 August 2021. Thirty-one days.

goeureka.io now serves a Thai-language Fun88 casino portal. The X account is deleted.

5. Trippki: the only one with a death certificate

Company dissolved 28 March 2023.

Hotel bookings paid in crypto, loyalty rewards in TRIP, reviews gated by a "Proof of Stay" model so only real guests could write them.

The ICO ran July to September 2018 with a 3,600 ETH soft cap and a 27,500 ETH hard cap. No result was ever announced. Two months later they published a post about "redesigning tokenomics" and partnering with third-party booking tech. Read that how you like.

Their last blog post went up 20 February 2020, weeks before travel stopped worldwide. Nothing was ever published again. A crypto OTA with no capital cushion and eighteen months of zero bookings had no path back.

TRIPPKI LTD, company number 10673507, registered to a house in Dorset, dissolved March 2023. Their final article was about whether European trains are greener than planes.

6. Travacoin: instant airline compensation

Institutional cover, no survival.

This one had everything a B2B pilot is supposed to have. The concept won third prize at IATA's World Passenger Symposium in Dubai, 2016. Wipro built the blockchain payment solution. IATA planned to test it. An FTI Consulting study claimed it would cut airline compensation costs 20 to 40%.

The product let airlines pay passengers instantly for delays and cancellations in digital currency instead of vouchers.

Then 2020 happened. Airlines stopped flying, faced a solvency crisis, and were being sued for refusing cash refunds. It was the worst conceivable moment to pitch a novel crypto refund instrument.

travacoin.com now returns a bare Microsoft IIS 404 page. The X account is deleted.

7. XcelTrip: the quiet exit

Never announced anything.

A crypto-only OTA out of Kathmandu and San Jose. Book hotels and flights with BTC, ETH, USDT or its own XLAB token. Part of a wider XcelLab ecosystem with a wallet and a marketplace.

There was no shutdown post. Content stopped in December 2024. The site was still up in November 2025. Today xceltrip.com and xcelpay.io have no DNS record at all. Somebody just stopped paying for the domains.

The ERC-20 contract still shows 101,810 holders. Roughly a hundred thousand wallets are holding a token whose company's domain name does not resolve.

8. Tripio: the biggest raise, the smallest footprint

Raised ~$19.2M. Dead by 2023.

More money than Winding Tree. Backers that read like a who's who of the 2018 Chinese crypto scene: Tron, F2Pool, Node Capital, INBlockchain. A decentralized booking marketplace out of Singapore with 5 billion TRIO minted.

It produced almost nothing anyone can point to.

Developer commits stopped in December 2022. TRIO stopped trading in January 2023. Today tripio.io returns NXDOMAIN, the domain was not even renewed, and an unrelated Hawaiian car-rental company briefly squatted the name before it went dark. The one artifact that survives is the ERC-20 contract itself, frozen on-chain with 18,275 holders and a market cap of exactly zero.

The lesson Winding Tree learned the hard way, Tripio never learned at all: raising more did not buy a single real booking.

9. Bee Token: the honest obituary

Raised ~$5M. Hacked during its own ICO. Shut down ~2019.

Beenest wanted to be a decentralized Airbnb. It is remembered for the wrong reason: on the final days of its January 2018 sale, phishers spoofed the team's mailing list and walked off with roughly $1M of contributors' ETH before the token even launched.

But the hack is not why it died. The team published one of the most honest shutdown notes in the sector, and it says exactly what Winding Tree's CEO said six years later: people showed up to invest, not to travel. BEE now trades at a fraction of a cent; beetoken.com is a memorial page, beenest.com does not resolve, and the X account is suspended.

Two independent teams, six years apart, wrote the same post-mortem. That is not a coincidence, it is the whole story of this sector in one sentence.

The one that did not fail: TravelbyBit

Worth separating out, because it gets miscounted as a casualty.

TravelbyBit put Bitcoin point-of-sale terminals into roughly 30 stores across Brisbane Airport in 2018. An entire international terminal taking crypto at the till. It raised $2.5M from Binance plus a Queensland government grant.

In May 2020 it merged into Travala, also Binance-backed. The entities wound down by 2022 and travelbybit.com now redirects to travala.com.

That is an exit, not a death. It is also the only physical-world footprint anyone in this sector ever built.

Its X handle now belongs to a stranger who registered it in June 2021.

What actually killed them

Across the whole cohort, the causes are not random.

Nobody needed a token to book a hotel. Travala proved you can run a crypto OTA at scale. It did it by being a good OTA that accepts crypto, not by making the token load-bearing. Every project here that made the token the product died.

The B2B ones needed the whole industry to move at once. Winding Tree and Travacoin both built things that only work if airlines and hotels coordinate. They do not coordinate. Winding Tree got the biggest names in aviation into pilots and still could not find one person at a hotel chain who could process a crypto payment.

COVID was the filter, not the cause. Trippki and Travacoin both died in 2020, but neither had a real business before it. Travel stopping for eighteen months just made the absence of revenue impossible to hide.

Regulated industries do not want public chains. Aeron's premise was correct and its market was closed. Flight records were never moving to Ethereum.

They came to invest, not to travel. This is the one line that recurs. Winding Tree's CEO wrote it in his 2024 post-mortem. Bee Token wrote it in its shutdown note in 2019. Nobody assembled a token holder base that actually wanted to book a hotel, they assembled a token holder base that wanted the token to go up. When it didn't, there was no user left underneath.

The domains keep ending up as casinos. Three of them now serve gambling sites: Pinktada, which the Wall Street Journal wrote about, points at an Indonesian slots page; GOeureka points at a Thai casino; RoomDAO now redirects to a Vietnamese betting portal. Dead crypto-travel domains have real search traffic, and the highest bidder for that traffic is almost always an offshore casino.

Raising more did not help. Tripio raised ~$19.2M and left less behind than teams that raised a tenth of that. Emphy raised only 6% of its target and shipped about as much. The size of the round had no relationship to whether anything got built.

The handful still standing did not survive because their tech was better. They survived because they sold something people wanted to buy, a working OTA, a live L1, an eSIM that connects, and the crypto part was a payment rail rather than the pitch.

The graveyard, in one table

The nine above got the long write-up because their stories are the most instructive. But they are not the whole cohort. Here is the full set of on-chain travel projects we could verify as dead or abandoned.

Winding Tree (LIF): $14.2M, Jul 2024. Got airline pilots live, found no one who could process crypto.

Tripio (TRIO): ~$19.2M, 2023. Biggest raise here, near-zero product; domain now NXDOMAIN.

Cool Cousin (CUZ): ~$10M, ~2019. Abandoned a working 500K-user guide app to chase a token.

Aeron (ARN): $5.7M, ~2022. Regulated aviation records were never moving to a public chain.

Bee Token / Beenest (BEE): ~$5M, ~2019. Phishing hack at ICO, then no users who wanted to travel.

Pinktada: ~$4.25M, Jul 2023. NFT hotel nights needed hotels and buyers that never showed.

GOeureka (GOT): 2 sales, Aug 2021. No product after ICO one; second life lasted 31 days.

Trippki (TRIP): cap unmet, Mar 2023. No cushion, 18 months of zero bookings, then COVID.

Travacoin: IATA-backed, ~2020. Pitched a crypto refund tool the moment airlines stopped flying.

XcelTrip (XLAB): ~2024. Quietly stopped paying for its own domains; ~100K holders stranded.

RoomDAO (RDC): ~$1.18M, ~2020. Small raise, stalled roadmap; domain now a Vietnamese casino.

Emphy (EPY): ~$0.45M, ~2020. Raised 6% of target, never shipped; domain gone.

Travelflex (TRF): own chain, ~2019. Promised its own DAG chain, delivered nothing.

Triip (TIIM): ~$0.7M, pivoted. Morphed into a "green metaverse" DAO that also stalled.

Aditus (ADI): ~2019. Luxury crypto-concierge with no traction (med confidence).

Dacsee (DACS): ~$31M, zombie. Blockchain ride-hailing lost to Grab; token dead, app limps (transport-adjacent).

TravelbyBit: $2.5M, 2022. Not a death, merged into Travala, the one clean exit.

Not a token, but worth a line: CryptoCribs, a bootstrapped crypto-Airbnb, voluntarily shut down in January 2021 and left an honest memorial page, the rare graceful exit.

The tally. Of roughly 20 on-chain travel projects we could verify from the 2017-2022 wave, about six or seven still genuinely operate, Travala, Dtravel, Roam, GET Protocol, LockTrip and Camino (the last on life support after its foundation began winding down). That is about one in four, and every additional dead token we did not chase down only pushes the survival rate lower. The card's "25% still survives" is, if anything, generous to the winners.

Domains now serving gambling sites: Pinktada (Indonesian slots), GOeureka (Thai casino), RoomDAO (Vietnamese betting). X handles deleted or suspended: GOeureka, Travacoin, Bee Token. Handles re-registered by strangers: Winding Tree, TravelbyBit.

Census verified 8 August 2026 from on-chain records, company registries, Wayback captures and live DNS. Confidence is high on every project shown except where marked. Not financial advice.

TL;DR: In Part 1 we mapped the machines. Part 2 maps the brains: 21 startups building the foundation models, training infrastructure and applied labs and farms behind physical AI. Together the two parts cover Bessemer's full 50-company map.

A robot is only as smart as the model that runs it and the tools that train it. If Part 1 was the bodies, humanoids, drones and self-driving vehicles, this is the intelligence layer: the models, data and simulation stacks that turn motors into autonomy. Using Bessemer's atlas again, we grouped 21 more private leaders into four categories. Every name links to the company.

Foundation Models

The models giving robots a mind of their own.

World Labs: Fei-Fei Li's spatial-intelligence startup building world models that generate and reason over interactive 3D worlds.

Physical Intelligence: vision-language-action robot foundation models (the pi series) built to control any robot on any task.

Skild AI: the Skild Brain, an omni-bodied general-purpose robot foundation model out of Pittsburgh.

Generalist: hardware-agnostic robot foundation models (GEN-0/GEN-1) for general-purpose manipulation.

Dyna Robotics: DYNA-1, a commercial-ready robot foundation model for high-dexterity task automation.

Perceptron AI: foundation models for real-time multimodal perception and embodied reasoning.

Eka Robotics: vision-force-action foundation models using tactile sensing and sim-to-real for dexterity.

Field AI: Field Foundation Models, a general robot brain for GPS-free autonomy in unstructured environments.

Robotics Infrastructure

The plumbing that trains, tests and ships the machines.

Foxglove: a data, visualization and observability platform for robotics and autonomous systems.

Point One Navigation: centimeter-level GNSS positioning for autonomous systems.

Voxel51: FiftyOne, a data-curation platform for computer-vision and physical-AI datasets.

Zeromatter: a simulation platform for robotics, autonomy and aerospace.

Theseus: a GPS-denied visual navigation system that lets drones fly without GPS.

Health & Life Sciences

Robots and AI moving into the lab and the operating room.

Periodic Labs: AI scientists plus autonomous labs to automate materials discovery, from ex-OpenAI and DeepMind researchers.

Medra: AI robots that automate repetitive wet-lab bench work.

Radical AI: a self-driving autonomous lab pairing AI with robotic experimentation for new materials.

Mendaera: Focalist, a handheld robotic system pairing robotics with real-time ultrasound for needle-guided procedures.

Climate & Agriculture

Autonomy pointed at fields, orchards and wildfires.

Carbon Robotics: LaserWeeder AI robots and tractor autonomy for herbicide-free farming.

Orchard: FruitScope, an AI vision system for precision crop and fruit management.

Seneca: autonomous AI drones that detect and suppress wildfires.

Upside Robotics: solar-powered autonomous field robots for precise crop nutrient application.

Why it matters for crypto

If Part 1's machines are the demand side of a machine economy, Part 2 is its supply chain, and much of it maps cleanly onto crypto's open-infrastructure thesis. Training these robot brains needs enormous compute and data, exactly what decentralized compute and DePIN data networks aim to coordinate and pay for permissionlessly. The data that teaches robots to see and act becomes an asset worth owning, pricing and verifying, and onchain provenance is one way to prove where a model's training data and outputs came from. As autonomy spreads from labs to farms to city streets, the pipes that move value and trust between all these agents are the same rails crypto is already building.

This completes our two-part Physical AI map. See Part 1 for the machines themselves.

Source: Bessemer Venture Partners, "50 Startups Transforming Industries With Physical AI." Categorization adapted by CryptoDiffer. Not financial advice.

TL;DR: Physical AI, robots and autonomous systems that can see, understand and act in the real world, is leaving the lab. Using Bessemer's new atlas as a base, we mapped 29 private leaders across four arenas. This is Part 1; Part 2 is coming.

"Embodied intelligence" is having its moment. After decades of promise, robots are moving out of research labs and onto factory floors, city streets, battlefields and into homes, driven by three forces at once: foundation-model breakthroughs, collapsing hardware costs, and real labor shortages. Bessemer projects up to 2.5 billion robots worldwide by 2035 and calls this one of the biggest technological inflections of the generation.

Below are 29 private companies (Waymo included) building the machines, grouped into four categories. Every name links to the company.

Industrial Robotics

The robots quietly rewiring warehouses, construction sites and factories.

Bedrock Robotics: retrofit autonomy kit that turns existing excavators and dozers into self-driving machines.

Cobot: Proxie, a collaborative mobile manipulator for material handling.

DroneDeploy: drone-and-robot reality capture to map and monitor jobsites.

Mind Robotics: AI-native industrial robots for dexterous manufacturing (a Rivian spinoff).

Mytra: software-defined warehouse robotics for moving and storing pallets.

Tutor Intelligence: robot arms as a service for pick-and-sort in manufacturing and logistics.

Sereact: Cortex, a vision-language-action model for hardware-agnostic robotic picking.

ANYbotics: ANYmal, a quadruped robot for autonomous industrial inspection.

RobCo: modular, no-code robot arms for small and mid-size manufacturers.

Consumer Robotics

Humanoids and home robots aiming for the living room.

Figure: general-purpose humanoids (Figure 03) running its own Helix AI.

Matic: a vision-based, privacy-first robot vacuum-and-mop for the home.

Sunday Robotics: Memo, a home robot that learns chores like dishes and laundry.

Zipline: autonomous delivery drones for food, groceries and medical supplies.

1X: NEO, a bipedal humanoid built for the home.

The Bot Company: a household robot for everyday chores, from Cruise co-founder Kyle Vogt.

Coco: autonomous electric sidewalk robots for urban last-mile delivery.

Autonomous Vehicles

The self-driving stacks moving people and freight.

Waymo: the robotaxi service built on the Waymo Driver.

Applied Intuition: simulation and software toolchain for autonomous vehicles and ADAS.

Wayve: end-to-end, mapless embodied-AI driving software.

Waabi: a generative-AI autonomy stack for self-driving trucks and robotaxis.

Nuro: a licensable Level 4 self-driving system, the Nuro Driver.

Defense

Autonomy for drones, ships and the battlefield.

Anduril: autonomous defense systems and the Lattice command platform.

Neros: low-cost, American-made FPV attack drones (Archer).

NODA AI: cross-vendor autonomy orchestration for military robot and drone fleets.

Auterion: AuterionOS, a drone operating system for defense and commercial fleets.

Saronic: autonomous unmanned surface vessels (sea drones) for the US Navy.

Shield AI: Hivemind, an AI pilot for autonomous drones and aircraft.

Skydio: AI drones for defense, public safety and infrastructure inspection.

Breaker Industries: Avalon, voice-controlled software orchestrating multi-domain robot teams.

Why it matters for crypto

Physical AI and crypto are converging faster than most think. As fleets of robots, drones and autonomous vehicles come online, they need to transact, coordinate and prove their work without humans in the loop, exactly the problem DePIN networks, machine-to-machine payments and onchain identity are built to solve. A world of billions of machines is also a world of billions of economic agents that can hold wallets, pay per task and settle in stablecoins. The hardware layer mapped above is where much of crypto's "agentic payments" and DePIN thesis will actually get tested.

This is Part 1 of our Physical AI map. Part 2, covering foundation models, agriculture, health and robotics infrastructure, is coming next.

Source: Bessemer Venture Partners, "50 Startups Transforming Industries With Physical AI." Categorization adapted by CryptoDiffer. Not financial advice.

TL;DR: 137 venture funds closed a combined $39.26B in Q2 2026, but the money is stacked at the top: 15 mega-funds took 69% of it. Crypto's share was $3.92B, and almost all of it is three familiar names.

Murph Capital's Q2 2026 recap frames the quarter as "concentration inside a contraction." The 15 platforms that cleared $500M captured 69% of all capital raised, while 76 emerging funds under $250M split just 13% between them. The headline number, $39.26B across 137 funds, hides how few hands it landed in.

The top of the table is generalist and AI-heavy. Sequoia Capital's $7B Expansion Fund leads, followed by Accel ($5B, Leaders Fund V) and Menlo Ventures ($3B). Those three alone account for $15B, more than a third of the entire quarter.

Crypto kept its seat at the table, but barely a broad one.

Dedicated crypto and Web3 vehicles raised $3.92B. That headline masks how narrow the base is: a16z crypto's $2.2B Fund V (the #4 fund overall, ahead of Benchmark and every biotech or defense fund on the board), Haun Ventures' $1B Fund II, and Framework Ventures' $400M together make up 92% of the whole category.

Murph's read is blunt: "Crypto's Q2 recovery is a story about 3 franchises re-upping, not broad LP re-entry into the sector." a16z crypto even passed on the AI diversification its peers chased, committing Fund V entirely to crypto founders.

Beyond crypto, Q2 funded a wide "hard" spread: physical and industrial AI at Eclipse Ventures ($1.3B), biopharma at Jeito Capital ($1.17B), European defense via E2D ($570M), healthtech at Lauxera Capital ($570M), and climate at Lightrock ($500M). It is capital positioning for physical infrastructure and sovereignty, the same real-world thesis that crypto's own RWA and DePIN narratives are chasing.

Full ranking, Q2 2026 (top 15)

1. Sequoia Capital: $7B, Expansion Fund (AI, multi-stage)

2. Accel: $5B, Leaders Fund V (late-stage AI)

3. Menlo Ventures: $3B, two vehicles (AI)

4. a16z crypto: $2.2B, Fund V (crypto, Web3)

5. Benchmark: $2B, two vehicles (early-stage, growth)

6. Eclipse Ventures: $1.3B, physical and industrial AI

7. Jeito Capital: $1.17B, Fund II (biopharma)

8. Haun Ventures: $1B, Fund II (crypto)

9. BAI Capital: $800M target (first close ~$600M), cross-regional

10. Luminous Ventures: $733M, China AI and robotics

11. 137 Ventures: $700M, growth

12. E2D: $570M, Fund I (European defense)

13. Lauxera Capital: $570M, Growth II (healthtech)

14. Lanchi Ventures: $560M, Fund IV (China AI)

15. Lightrock: $500M, Accelerate7 (climate)

What it means for crypto: The funding class is normalizing crypto into the same table as biotech, defense and climate, a mature asset class competing on merit rather than hype. But this quarter's crypto capital is a few franchises re-upping, not a wave of new LPs. When a16z crypto, Haun and Framework deploy that $3.92B, the projects raising in 2026 and 2027 will feel it.

Data: Murph Capital, Q2 2026 recap. Fund sizes are self-reported closes and may revise. Not financial advice.

One of South Korea's biggest crypto exchanges is swapping the memecoin fast lane for something far more buttoned-up: a listing on the stock market.

Bithumb is targeting a 2028 IPO on Korea's KOSDAQ, with a possible Nasdaq encore later to court international investors. It's a slow burn, and the groundwork is already well underway.

Back in July 2025, Bithumb split itself in two: Bithumb Korea (the exchange that will actually list) and Bithumb A (its asset-management arm), to keep the corporate story clean for future shareholders. Since then it has been grinding through the unglamorous but necessary prep: upgrading internal controls, moving to K-IFRS accounting, and lining up Samsung Securities as underwriter.

A few asterisks, though. The 2028 date is a target, not a promise, and Bithumb already shelved an earlier 2025 attempt when markets turned choppy. No valuation has been floated yet. And to be clear, this is an equity IPO, not a token launch, so there is no "Bithumb airdrop" to hunt for.

If it lands, Bithumb would be among the first Korean crypto exchanges to ring the opening bell, a notable moment for one of the world's most active crypto markets. Watch this space.

Not financial advice. Based on reporting (The Block) and Bithumb's stated plans; the timeline is a target and may change.

On Monday, July 27, 2026, a company most crypto traders had never heard of became the most valuable listed firm in mainland China. ChangXin Memory Technologies, known as CXMT, priced its Shanghai IPO at 8.66 yuan and closed its first day up about 466%, briefly worth more than Intel. It raised $8.6 billion, the biggest mainland listing since Agricultural Bank of China in 2010 and Asia's largest IPO of the year. And here is the part that matters for this audience: crypto markets had already priced it, on pre-IPO futures, before a single share traded on the exchange.

Here is the full breakdown.

What is CXMT?

CXMT is a memory chipmaker founded in 2016 in Hefei, Anhui Province, to reduce China's reliance on imported memory. It is now the country's largest maker of DRAM, the memory inside phones, PCs, servers and AI systems, and the fourth-largest DRAM producer in the world behind Samsung, SK Hynix and Micron. Its global DRAM share has climbed from under 4% to roughly 8% in a year, and it is the only Chinese company producing modern DDR5, LPDDR5 and LPDDR5X at scale.

The IPO that broke records

CXMT listed on Shanghai's tech-focused STAR Market under ticker 688825. It priced shares at 8.66 yuan and raised 57.92 billion yuan, about $8.6 billion, the largest semiconductor offering in mainland China on record. Demand was staggering: retail investors oversubscribed their tranche 212 times, submitting 9.4 million orders and freezing more than 800 billion yuan of cash. The institutional tranche was oversubscribed more than 500 times.

The debut: 466% and a new number one

Shares opened at 49.50 yuan against the 8.66 yuan offer price and closed the day up around 466%. That gave CXMT a market capitalization of about 3.28 trillion yuan, roughly $485 billion, making it the most valuable company listed on China's mainland exchanges, ahead of ICBC and above Intel by market value.

The financials behind the frenzy

This was not a story stock with no revenue. In the first quarter of 2026, CXMT reported revenue of about $7.1 billion (50.8 billion yuan), up a remarkable 719% year over year, with net profit of roughly $3.5 billion. For the first half of 2026, the company guided to revenue of 110 to 120 billion yuan and net profit of 50 to 57 billion yuan. The AI memory boom, which has pushed DDR5 and HBM prices sharply higher, is landing directly on CXMT's income statement.

A decade of state money, and Hefei's giant bet

CXMT did not appear overnight. It raised capital across nine rounds over ten years, and state-owned shareholders held 36.29% before the IPO, including China's flagship National Integrated Circuit Industry Investment Fund (the "Big Fund"), plus Hefei and Anhui government vehicles. Hefei alone poured more than 30 billion yuan into the project over a decade and kept investing through 36.65 billion yuan of accumulated losses, never exiting. Monday's debut turned that patient, counter-cyclical bet into trillions of yuan of paper value. Chairman Zhu Yiming, who earlier founded flash-memory firm GigaDevice, has led CXMT since 2018.

Who actually buys CXMT memory

According to its IPO prospectus, CXMT's customers include Tencent, Alibaba Cloud, ByteDance, Lenovo and Xiaomi. Tencent alone signed a server-DRAM supply deal worth around $3 billion. CXMT memory already ships inside Lenovo ThinkPad and Huawei MateBook devices, and Western brands including HP, Dell, Acer and Asus have evaluated or adopted modules built on its chips as it entered global PC supply chains.

Built on a homegrown tool chain

Part of what makes CXMT strategically important is its supply chain. Domestic equipment now accounts for roughly 40% to 50% of the tools on its production lines, supplied by Chinese vendors such as NAURA, AMEC, Piotech, Hwatsing, ACM Research and Kingsemi. The one hard ceiling remains lithography, still dominated by ASML and Nikon, where China's domestic alternative has only a sliver of the market. That equipment gap, reinforced by US export controls, is the single biggest constraint on how fast CXMT can scale.

Chasing the big three

CXMT sits fourth in DRAM, but research suggests its capacity is closing in on Micron, which would put China on track to become the world's second-largest DRAM producer by capacity. It currently runs about 300,000 wafers per month across three fabs, with two new plants targeting 600,000 wafers per month. It is also moving into high-bandwidth memory, targeting 55,000 HBM wafer starts per month by 2027 and 100,000 by 2028, aiming directly at the AI accelerator supply chain that Samsung and SK Hynix dominate today.

The crypto angle: priced before it listed

For crypto traders, the most interesting detail is timing. CXMT traded on pre-IPO futures for weeks before it listed, making it the first major Asian company stress-tested by crypto-native pre-IPO markets. It now trades as a perpetual, CXMTUSDT, letting anyone take a position on China's memory champion without a Chinese brokerage account. It is a clean example of the same convergence we have covered elsewhere: tokenized stocks, 24/7 trading, and crypto rails pricing traditional assets faster than the traditional market can.

The risks

None of this is a one-way bet. US export controls cap CXMT's access to the most advanced chipmaking equipment, a hard limit on scaling. Its valuation, about $485 billion on roughly 8% market share, prices in years of flawless execution. And a debut that runs 466% in a day tends to give some of it back. This is a strategically vital company and a volatile stock at the same time.

The takeaway: CXMT is the clearest sign yet that China's push for memory self-sufficiency has produced a genuine global contender, and that crypto markets are increasingly where price discovery starts. Whether the stock holds its trillion-yuan valuation is a separate question from whether the company matters. It clearly does.

CXMT IPO: quick FAQ

What is CXMT's ticker? 688825 on the Shanghai STAR Market. It also trades as the CXMTUSDT perpetual in crypto.

How much did CXMT raise? About $8.6 billion (57.92 billion yuan), the biggest mainland Chinese IPO since 2010.

How much did CXMT stock rise on debut? Around 466% on July 27, 2026, reaching a market cap near $485 billion.

What does CXMT make? DRAM memory chips, including DDR5, LPDDR5 and LPDDR5X, and it is developing HBM for AI systems.

Enjoyed this? The Differ delivers crypto's week in 3 minutes, every Sunday. → Subscribe

Nothing here is financial advice. Do your own research.

Prediction markets had their breakout year — Polymarket and Kalshi cleared record volumes, and the category crossed $20B in monthly turnover. But the loudest names are the venues. The real land-grab is quieter: the startups building the rails underneath — data, terminals, matching engines and credit — and most are still pre-token.

Wintermute put it bluntly this month: the prediction-market boom has an infrastructure problem. Fragmented data, slow tooling, no professional-grade APIs. So we went hunting for the teams fixing it. Ground rules: we skipped the big traded names (UMA, Chainlink) and the venues themselves, and kept only fresh, mostly pre-token builders. Inclusion isn't an endorsement — do your own research.

Here's the infrastructure map 👇

Data / API

1. Oddpool — Aggregates odds, orderbooks and arbitrage across every prediction venue into one searchable data layer for quants. Y Combinator (S26).

2. PredictionTalk — A prediction-markets hub tracking TVL, volume, fees and oracles across every major venue, updated continuously. Independent.

3. Adjacent — A data-and-news API surfacing 40,000+ markets and prediction indices, turning odds shifts into breaking news. Pre-seed, 2026.

4. Marketlens — Tick-level Polymarket orderbook history plus a Python backtesting SDK for quants building and testing real strategies. Independent.

Terminals / Execution

5. Kairos — Unified execution-and-intelligence terminal aggregating data, news and low-latency trades across venues. $2.5M, a16z crypto.

6. Verso — Bloomberg-style terminal with real-time data, AI news scoring and smart order routing across Polymarket and Kalshi. Y Combinator backed.

7. Fireplace — Institutional terminal aggregating markets, liquidity and execution, auto-routing orders to the deepest, best-priced venue. $1.5M pre-seed, Frachtis.

8. Converge — The first unified terminal letting traders watch and act across Polymarket, Kalshi and Limitless from one screen. Independent.

Protocols / Credit / Infra

9. Pascal — Core infrastructure powering the launch and operation of prediction-market platforms. $9M Series A, Union Square Ventures.

10. Kuest — Open-source protocol to launch and run your own prediction market. Independent · open-source.

11. OrderbookTrade — High-performance CLOB matching engine with on-chain settlement, launching markets instantly. Independent.

12. Dimes — Embedded credit letting any app offer leveraged prediction-market exposure via API. Independent.

The takeaway: everyone is trading on Polymarket. The smarter bet may be on who's building the layer beneath it — and unlike the venues, most of these are still early and pre-token.

Enjoyed this? The Differ delivers crypto's week in 3 minutes, every Sunday. → Subscribe

Nothing here is financial advice. Early-stage projects carry high risk — DYOR.

No matter what the market's doing — green, red, sideways — new projects keep launching every week. And the best window to get in is the earliest one: before the token, before the crowd, while the door's still open.

We tracked down 18 active waitlists you can join right now — from prediction markets and DEXs to crypto cards and privacy AI. One rule first: a real waitlist only ever needs an email or a social handle. Never connect a wallet, sign a transaction, or share a seed phrase to "join" one. With that out of the way — here's the radar.

1. Arcus — The native DEX for Robinhood Chain, bringing early spot trading to the fastest-launching L2 in history. Website

2. blink — A high-performance exchange for spot and perpetual markets, built for fair access and blockchain-native settlement. Website

3. Clawx — An agentic prediction market on Avalanche, where AI agents trade real-world outcomes on your behalf. Website

4. Daily Fun — Turns market speculation into a social, game-like experience for everyday traders. Website

5. Hudi — Trade Asian stock perpetuals — Nikkei, KOSPI, Hang Seng names — with up to 25x leverage. Website

6. KOR Protocol — An on-chain platform for tokenizing, managing and monetizing digital intellectual property, built on Base. (Raised $7.5M) Website

7. Koto Trade — A trading platform where the first 250 users on the waitlist receive a funded account at launch. Website

8. Onyx — A social sports prediction platform to trade positions on game outcomes. (Raised $20M from Payward) Website

9. Pinnaccle — A wallet-native trading workspace on Canton — swaps, limit orders, DCA and stop-loss in one place. Website

10. Proof — A new prediction-market venue for trading on real-world events and outcomes. Website

11. Pulsar Money — Free USD and EUR accounts, 0% stablecoin funding, and a virtual dollar card for global spending. Website

12. Pyre — A crypto card for everyday spending, paid straight from your on-chain balance. (Supported by Sophon) Website

13. Rocket Markets — The first time-based prediction yield market — earn yield on where outcomes are heading. Website

14. Shift DeFi — Non-custodial DeFi asset management targeting up to 15% yield via structured, risk-managed strategies. Website

15. Sovra — A self-custodial USDC platform for saving, earning and spending digital dollars anywhere. (Raised $2M) Website

16. Startale Card — A crypto spending card from Startale, with early-access registration now open. (Startale raised $70M) Website

17. Tplus — A decentralized trading platform combining on-chain settlement with fast off-chain order matching. Website

18. Wisp — Privacy-focused AI where your data is never stored or used for training. Fully owned by the Lido DAO. Website

───

Want the next Early Access Radar in your inbox? Subscribe to The Diff — free, weekly, 3-minute reads. Subscribe →

Not financial advice. Early-stage projects carry real risk — do your own research, and never share wallet credentials.

No matter what the market's doing — green, red, sideways — new projects keep launching every week. And the best window to get involved is the earliest one: before the token, before the crowd, while the door's still open.

So we rounded up 15 fresh projects with open waitlists into one Early Access Radar — AI trading, perp DEXs, prediction markets, crypto cards. A few minutes each, all in one place, so you don't have to go hunting.

One rule first: these are early-stage and unproven. A waitlist is a cheap option, not a promise — do your own research, and never connect a wallet or send funds to something you haven't checked. With that said, here's the board 👇

1. Aurus — Trading platform on Solana with lower fees and wallet cost-basis tracking. Website

2. Barter — DeFi infrastructure for liquidity and capital efficiency across networks. Raised $3M (Maven 11). Website

3. Boost (RabbitHole) — Protocol to deploy, manage and distribute on-chain token incentives. Website

4. Fact Machine — Opinion market where users bet on which view wins a majority vote. Website

5. Flip — AI financial assistant built right into iMessage. Raised $1.4M (a16z). Website

6. Horizon — No-code platform to build, backtest and automate trading strategies. Website

7. Hypernova — On-chain prop trading on Arbitrum with instant smart-contract payouts. Raised $3M (Lemniscap). Website

8. HyperSignals — Social on-chain analytics and copy-trading, built on Hyperliquid. Website

9. JTX — Self-custody trading + perps on Solana, from the team behind Jito. Website

10. Kairos — Execution and intelligence layer for prediction markets (Kalshi + Polymarket). Raised $2.5M (a16z). Website

11. MNX — Futures DEX for AI valuations, stocks, compute prices and prediction markets. Raised $6.4M. Website

12. Nof1 — AI research lab behind Alpha Arena, where LLMs trade live capital. Raised $15M (Sui Group). Website

13. Pear — Social-centric prediction-market platform, multiple venues in one interface. Website

14. Slush — Slush Card: a crypto debit card for the Sui ecosystem, by Mysten Labs. Website

15. Superluminal — Perpetual DEX on Fogo Chain using a Dual-Flow Batch Auction. Website

Want the next Radar first?

The Differ sends takes like this every week — the signal, the one number, the move, in ~3 minutes. Free. Subscribe free →

None of this is financial advice. Early-stage projects carry real risk — DYOR.

The first half of 2026 confirmed what the year's TGE pipeline promised: token launches have shifted away from early-stage experiments toward platforms with live products, measurable volume, and real user bases. The summer slate continues that trend, concentrated heavily on on-chain derivatives, AI, and DePIN infrastructure, and the next wave of consumer-facing applications.

This guide breaks down 30 of the most anticipated TGEs heading into summer 2026 by category, covering what each project does, what to expect from its token launch, and where it sits in the broader market.

What Is a TGE and Why the Summer Slate Matters 

A Token Generation Event (TGE) is the point at which a project issues its native token and distributes it to investors, the community, or both. It marks a project's transition from a centralized product toward a token-governed ecosystem, and it is typically the first moment the market can price a network directly.

The summer 2026 lineup is dominated by trading infrastructure. More than a third of the projects below operate in the DEX and perpetuals category, reflecting the continued migration of derivatives volume on-chain. Alongside them, AI and DePIN networks, GameFi platforms, and a small group of Layer 1 and Layer 2 chains round out a pipeline weighted toward products that already process activity rather than concepts still in development.

Layer 1, Layer 2, and Infrastructure TGEs

Base (Layer 2) — Coinbase's Ethereum L2 remains one of the most active chains in crypto by daily transactions. No official TGE date has been confirmed, but expectations for a token remain among the highest in the L2 category given the size of its on-chain user base.

Ink (Layer 2) — Kraken's L2 network has been building ecosystem momentum since launch. A token would extend governance to its user community and formalize ecosystem incentives.

Teneo (Layer 2) — A Layer 2 network expected to TGE as it scales its ecosystem, distributing tokens to early users and builders.

Arc (Layer 1) — A new Layer 1 entrant targeting high-performance execution for next-generation applications. A token launch would anchor its validator and ecosystem incentive structure.

Aligned (Infrastructure) — A verification and proof infrastructure layer designed to serve other networks and applications. Its token is expected to underpin the economic security of its proving system.

Arcium (Privacy) — A privacy-focused computation network enabling confidential on-chain processing. A TGE would tie network usage and node operation to its token economy.

DEX, Perps and Trading Infrastructure TGEs

On-chain derivatives are the dominant theme of the summer pipeline, with a cluster of perpetuals venues expected to formalize their token economies.

GRVT (DEX/Perps) — A hybrid ZK exchange combining centralized-style performance with on-chain settlement. It is one of the larger names in the perp DEX category by trading volume.

Pacifica (DEX/Perps) — A perpetuals DEX that has drawn significant attention as Solana-based derivatives activity has grown. A token launch would reward early traders and liquidity providers.

Hibachi (DEX/Perps) — A perpetuals exchange focused on performance and capital efficiency, expected to TGE as it scales open interest.

Extended (DEX/Perps) — A derivatives platform positioned in the competitive perp DEX field, with a token anticipated to support fee and incentive mechanics.

StandX (DEX/Perps) — A perpetuals venue building toward a token launch tied to trading activity and ecosystem rewards.

Nado (DEX/Perps) — A perpetuals DEX expected to distribute tokens to early users as it formalizes its market structure.

BULK (DEX/Perps) — A derivatives platform in the perps category, with a TGE anticipated alongside its growth phase.

Strata (DeFi/Perps) — A protocol combining DeFi primitives with perpetuals exposure, expected to launch a token covering both sides of its product.

Everything (DeFi/Perps) — A platform spanning DeFi and perpetuals products, targeting users who want combined spot and derivatives access.

Satsuma (DeFi/DEX) — A DeFi and DEX protocol expected to TGE as it builds liquidity and trading depth.

AI, DePIN and DeFAI TGEs

The intersection of AI, decentralized physical infrastructure, and on-chain finance continues to produce a distinct token category.

Krain AI (AI/Infra) — An AI infrastructure project focused on model deployment and optimization pipelines, expected to launch a token as it scales its platform.

PIN AI (AI/Service) — A personal AI protocol that stores and monetizes user context on-chain, representing a data-ownership approach to AI services.

Mawari (DePIN/AI) — A spatial computing infrastructure network combining DePIN hardware incentives with AI rendering for immersive applications.

YOM (GameFi/DePIN) — A network bridging gaming and decentralized physical infrastructure, using token incentives to coordinate compute and rendering resources for games.

ChimpX (DeFAI) — A project in the emerging DeFAI category, applying AI-driven automation to on-chain financial strategies.

GameFi, Marketplace and Consumer App TGEs

Cambria (GameFi) — An on-chain game expected to TGE as it builds its player base and in-game economy.

MagicBlock (GameFi) — A gaming infrastructure protocol providing real-time on-chain execution for games, expected to launch a token tied to network usage.

Spaace (Marketplace) — A marketplace protocol expected to distribute tokens to users and participants as it scales transaction volume.

Printr (MemePad) — A token launch and memecoin issuance platform, with a TGE anticipated as part of its ecosystem growth.

Veera (Neobank) — A crypto-native neobank expected to launch a token as it expands into decentralized financial services.

RWA, Lending, Predictions, Security and Data

STRATO (Lending/RWA) — A collateralized debt position and basis-trading infrastructure protocol spanning lending and real-world assets.

Zetarium (Predictions) — A prediction market protocol expected to launch a token as the sector expands beyond its incumbents.

Quip (Security) — A decentralized security protocol focused on on-chain threat detection and smart contract monitoring.

Tea (Data Service) — A data service protocol expected to TGE as it tokenizes access to its infrastructure.

Key Trends Across the Summer 2026 Pipeline

Several patterns stand out across this slate.

Derivatives dominate. More on-chain perpetuals venues are launching tokens than any other category. This reflects the continued shift of derivatives volume from centralized exchanges to on-chain venues, and the competition among platforms to lock in liquidity and traders ahead of their TGEs.

AI and DePIN keep converging. Projects such as Mawari and YOM sit at the intersection of AI workloads and physical hardware incentives, while DeFAI entrants like ChimpX apply AI automation to financial strategies. The category continues to mature into a distinct asset class.

Infrastructure remains anticipated but undated. The largest potential launches, including Base, carry the highest expectations but no confirmed timelines. Token schedules in crypto remain aspirational and subject to market conditions, regulatory developments, and product readiness.

Consumer reach is broadening. Neobanks, marketplaces, and GameFi platforms in this slate are oriented toward end users rather than purely on-chain-native audiences, continuing the move of token models toward products built for everyday use.

Conclusion

The summer 2026 TGE pipeline is concentrated where on-chain activity is concentrated: derivatives trading, AI and DePIN infrastructure, and consumer applications with live products. Rather than a wave of untested protocols, this slate is led by platforms already processing volume and serving users.

As always, whether these launches deliver depends on token design, distribution fairness, and genuine governance utility. The pipeline is real and the categories are clearly defined, but timelines remain the least predictable variable in any TGE.