quote: Seeker Season 2 by @solanamobile is live and we shipped exclusive features for Seeker users
• open Suitcases and get rewards
• collect Travel Points
• use them for bookings
Travel with @NOMADZxyz on Seeker https://x.com/solanamobile/status/2097077264055718235 | Seeker Season takes off with @NOMADZxyz ✈️
Book 1M+ stays with stablecoins, better rates, and rewards built around travel.
Seekers get a free Suitcase NFT, a 30% XP boost, and early access to Nomadz Travel Points.
Available now on the Solana dApp Store.
0·-Neutral
n
news9/4news
Stellar Q2 Token Holder Report
Executive Summary
Q2 2026 brought the institutional adoption Stellar spent a decade building toward. On May 27, the Depository Trust & Clearing Corporation announced plans to connect the Depository Trust Company's tokenization service to Stellar, a venue that custodies and services more than $114 trillion in securities. Tokenized assets are expected to become available on Stellar in the first half of 2027. In the interim, DTCC and the Stellar Development Foundation will evaluate use cases across constituents of the Russell 1000, ETFs tracking major indices, and US Treasury bills, bonds, and notes.
What the numbers do reflect is asset growth. Tokenized real-world assets closed Q2 at $3.00 billion, up from $1.50 billion at the end of Q1, and the network passed three separate billion-dollar marks inside a single calendar year. Payment volume recovered to $16.43 billion, up 68.8% from Q1's $9.73 billion. Stablecoin balances grew 16.0% to $346 million.
Underneath the asset growth, the composition of network activity kept shifting. Stellar settles two kinds of activity on a single ledger: classic operations (the original payment, orderbook, and trustline primitives) and Soroban (the smart contract environment added in a February 2024 protocol upgrade). Both are validated by the same set of validators, so the split between them describes where activity settles rather than a choice users make. Soroban accounted for 41.3% of all Stellar transactions in Q2, up from 38.8% in Q1 and 25.3% three quarters earlier. Soroban transactions grew 14.3% quarter over quarter against 3.1% for classic transactions, and Soroban now carries 94.4% of network fees and 82.4% of onchain trading volume. Stellar is measurably deepening into a DeFi ecosystem alongside its payments business.
The upgrade cadence held. Protocol 26, "Yardstick," went live on mainnet on May 6 after an April 16 testnet launch, introducing CAP-77 Quorum Freeze, a protocol-native onchain account freeze mechanism that lets validators contain compromised ledger keys through federated consensus. Protocol 27, "Zipper," reached testnet on June 18 and activated on mainnet on July 8, shortly after the quarter closed.
Network fees fell 18.2% to $90,440 as the cost of using the network continued to come down by design. Whisk, Protocol 23, went live on mainnet in September 2025 and restructured rent and resource pricing, and the network's soroban limit parameters cap what a single transaction can consume, holding per-transaction cost down as usage grows. Fees have fallen in every quarter since, while transaction counts have risen in each of them. Daily active addresses averaged 63,442, down 10.2% from Q1. Address counts and asset growth are not expected to track each other closely on this network, because tokenized Treasuries and similar instruments are bought and held rather than transacted daily.
Financials
Stellar is a network supported by a nonprofit foundation, and its financial picture should be read that way. The Stellar Development Foundation exists to grow a network that is open to everyone, and there is no protocol revenue in the sense that a fee-capturing application has revenue.
Total transaction fees for Q2 2026 were $90,440, equivalent to 507,780 XLM, averaging $994 per day. That is a decline of 18.2% from Q1's $110,519, and it extends a four-quarter contraction from $553,931 in Q3 2025, through $268,161 in Q4 2025 and $110,519 in Q1 2026. Against 444.5 million transactions in the quarter, the implied average fee was roughly $0.0002.
The decline is a product decision with a specific mechanism behind it. Whisk, Protocol 23, was activated to restructure rent and read costs so that contract execution became materially cheaper. Alongside it, the limits set by the Stellar Limits Proposal cap the resources a single transaction can consume, which holds per-transaction cost down as usage grows. Fees have fallen in every quarter since Whisk went live, while transaction counts have risen in each of those quarters. Cheap access is the objective. The fee line is where it’s demonstrated.
The composition of those fees is the more informative figure. Soroban accounted for 94.4% of network fees in Q2 while handling 41.3% of transactions. Contract execution consumes metered resources, meaning instructions, ledger reads and writes, and rent for persistent state, where classic payment operations pay a flat base fee. As contract activity grows as a share of the network, fees concentrate there, and classic payments contribute very little to the pool.
Stellar's transaction fees collect in a locked fee pool. They are not burned, not distributed to validators, and not spendable by the Foundation. There is no staking yield, because the Stellar Consensus Protocol is not proof-of-stake, and there is no buyback or burn mechanism. Fees accumulate outside circulation.
XLM circulating supply closed the quarter at 33.98 billion, up 2.8% from 33.04 billion at the end of Q1. Stellar removed its inflation mechanism by community protocol upgrade in 2019, so the increase is distribution from Foundation-held balances rather than new issuance. Either way, 0.93 billion additional XLM reached the market during the quarter, and the proportional claim of existing holders fell accordingly. Q2 was the fastest quarter of distribution in the past year, ahead of 0.66 billion in Q1 and 0.48 billion in Q4 2025.
The Foundation's position is disclosed onchain across labeled accounts covering development, ecosystem growth, product and innovation, and assets and liquidity. Deducting circulating supply, the locked fee pool, and the upgrade reserve from the 50.00 billion total leaves approximately 15.76 billion XLM in SDF mandate accounts at June 30, equal to 46% of circulating supply and 32% of total supply. At the Q2 distribution rate that balance represents several years of potential supply. The holdings are large, identifiable, and publicly trackable. Scale and transparency are separate facts about them.
Network Activity & Payments
Stellar processed 444.5 million transactions in Q2 2026, up 7.4% from Q1, averaging 4.89 million per day. The share that failed fell to 26.0%, a fourth consecutive quarterly improvement from 36.0% in Q3 2025. Operations grew faster, reaching 1.10 billion, up 29.4% quarter over quarter. Operations are the unit of user intent, since each one is a discrete instruction a sender asked the network to carry out, while resource utilization is the under-the-hood quantity that sets what a transaction costs.
The rise in operations per transaction came entirely from the classic side of the ledger. Classic transactions carried 3.50 operations each in Q2 against 2.72 in Q1, a 29% increase on a classic transaction count that grew only 3.1%. Soroban transactions carry exactly one operation each, in this quarter and every prior one. Senders, meaning developers and applications, package multiple operations into a single classic transaction envelope before submitting it, and they did more of that this quarter. Because a Soroban envelope carries one operation regardless of how much computation it performs, operation counts cannot be used to compare the scale of work in the two environments.
Address activity declined. Daily active addresses averaged 63,442 in Q2 against 70,632 in Q1, down 10.2%, and below the 75,283 of Q3 2025, the high of the five-quarter window.
Daily address counts measure participants who transact on a given day, which is a poor proxy for holdings in assets that are bought and held. Tokenized Treasuries, corporate credit, and similar instruments generate little recurring transaction activity by design. Asset growth and daily address counts are therefore not expected to move together on this network, and the gap between them this quarter should not be read as either confirming or contradicting the other.
Payment volume recovered to $16.43 billion, up 68.8% from $9.73 billion in Q1. USDC carried 56.3% of that volume, XLM 36.3%, other stablecoins 4.2%, tokenized real-world asset transfers 2.7%, and all remaining assets 0.6%.
The distribution matters more than the total. Median daily payment volume was $159 million while the mean was $181 million and the single largest day reached $735 million. Three days account for a disproportionate share of the quarter. On May 29 the network settled $735 million, of which $401 million was tokenized real-world asset transfers, a single institutional settlement rather than recurring throughput. On June 29 and June 30, volumes of $731 million and $715 million were almost entirely USDC. Underneath those spikes the network ran a stable base of roughly $150 million to $180 million a day, and that base is the better guide to what recurs.
Disbursements through the Stellar Disbursement Platform totaled $0.6 million for the quarter across 2,050 payments. These are programmatic aid and remittance flows, and they are reported separately from organic payment activity rather than blended into it.
Stellar's stablecoin throughput can be measured two ways, and the answers differ by more than a factor of two, so the basis matters whenever the figure is quoted. Counting payment operations, which is what a user would recognize as sending money, stablecoin volume was $9.93 billion in Q2, up 80.2% quarter over quarter. Counting every stablecoin transfer across all operation types, including DEX trades and contract movements that are not user payments, the Foundation reports $11.4 billion, up 72%.
This report uses the payment-operation figure throughout, because it corresponds to economic settlement rather than internal contract mechanics.
Tokenized Assets & Stablecoins
Real-world assets
Tokenized real-world assets on Stellar closed Q2 2026 at $3.00 billion, up 100.0% from $1.50 billion at the end of Q1. Over five quarters the total grew roughly sixfold, from $493 million at the end of Q2 2025. The network crossed $1 billion in January, $2 billion in April, and $3 billion in June. The pace is the notable part, not the level. Every figure in this section counts tokenized real-world assets and excludes stablecoins.
By asset class at quarter close, US Treasury debt is the anchor at $1,203 million across 16 instruments. Active strategies follow at $560 million across 6, corporate credit at $513 million across 4, and non-US government debt at $487 million across 6. Real estate accounts for $166 million across 18 instruments and tokenized stocks $26 million across 6. Four smaller classes hold the remaining $42 million between them: private equity at $23 million, asset-backed credit at $16 million, and commodities and diversified credit at under $2 million each. The concentration in government and corporate fixed income is what makes the DTCC scope, which centers on Treasuries and index constituents, a natural extension of what is already there.
Composition above is a single-quarter position, not a trend. The subclass taxonomy was recut twice during the past year, and categories split rather than renamed, so subclass balances are not comparable across quarters even though the totals are.
Issuer concentration is the risk that deserves naming. Thirteen issuers held tokenized assets on Stellar at quarter close. The top four held 92.5% of the total. Spiko accounted for $1,089 million, or 36.3%, across 9 instruments. Franklin Templeton held $596 million (19.9%) across 4, Realiz $558 million (18.6%) across 3, and Ondo $530 million (17.7%) in a single instrument. Ranked by issuer, Franklin Templeton is second to Spiko. Ranked by individual asset, Franklin Templeton's BENJI sits behind Ondo's USDY, so descriptions of BENJI as the largest or second-largest tokenized asset on Stellar depend on which of those two bases is meant. The practical implication is that the headline figure is presently a function of four issuers' commercial decisions, and the migration of any one of them would be visible in the total immediately.
Stablecoins and wrappers
Stablecoin balances closed Q2 at $346 million, up 16.0% from $298 million at the end of Q1. USDC dominates at $258 million held across 643,185 accounts. Two issuances are new this quarter: MoneyGram's MGUSD at $25.0 million across 4 holding accounts, and Figure's YLDS at $25.0 million across 7. Supply in both sits with issuers rather than distributed users. Holder counts would need to rise before those balances represent retail circulation. Among the established issuances, SG Forge's EURCV held $17.1 million, PYUSD $7.9 million, Novatti's AUDD $3.8 million, and Circle's EURC $3.1 million across 4,582 accounts.
Stablecoin participation was flat. Daily senders averaged 24,331 in Q2 against 24,442 in Q1, and daily receivers 22,484 against 22,357, both within a percentage point of the prior quarter. These counts cover classic Stellar accounts only; activity held in contracts, liquidity pools, and claimable balances is excluded. Set against 16.0% balance growth and 80.2% growth in stablecoin payment volume, the same number of participants moved more value.
Regulated wrapper exposure remains small. Four spot exchange-traded products referencing XLM held $69.9 million at quarter close, taking in $2.2 million of net inflows against effectively flat flows in Q1. The category is immaterial to the network's economics at this size.
Protocol Analysis
Platform migration
The clearest structural trend on Stellar is the growth of Soroban contract execution alongside classic payment operations. Both run on the same ledger and are secured by the same validators. Soroban handled 41.3% of all network transactions in Q2, up from 38.8% in Q1 and 25.3% in Q3 2025. The growth differential drives the mix: Soroban transactions rose 14.3% quarter over quarter while classic transactions rose 3.1%. On economic measures the shift is further along than the transaction count suggests, with Soroban carrying 94.4% of network fees and 82.4% of onchain trading volume.
Onchain trading volume totaled $2.78 billion in Q2, up 54.8% from $1.80 billion in Q1. Soroban venues carried $2.29 billion of that, the classic orderbook $454 million, and classic automated market maker pools $36 million. Trade count grew faster still, reaching 63.8 million against 37.3 million in Q1.
DeFi total value locked closed the quarter at $196 million, up 18.4% from $166 million at the end of Q1. The intra-quarter path was not smooth. TVL peaked at $243 million on May 30 before retracing, and touched a low of $158 million on May 23.
Consensus and security
The validator set expanded over the quarter, from 87 active validators at the close of Q1 to 94 at the close of Q2. The categories are nested, not additive. Of the 94 active validators at quarter close, 70 were full validators that also publish history archives, and 21 of those sat in the tier-one quorum set that anchors network consensus. Set alongside the Foundation's publicly labeled mandate accounts, this is the concrete material for assessing decentralization: the validator count is growing, tier-one membership is small and identifiable, and Foundation holdings are trackable. Stellar's consensus does not use stake weighting, so validator influence is a function of quorum configuration rather than token holdings.
Protocol 26, "Yardstick," reached mainnet on May 6, 2026 following an April 16 testnet deployment. Its headline capability, CAP-77 Quorum Freeze, is the first protocol-native onchain account freeze mechanism on a major layer 1, allowing validators to freeze compromised ledger keys through federated consensus within minutes rather than hours. It answers a pattern of exploits across the industry, where the absence of formalized emergency tooling forced ad-hoc coordination. A February 2026 DeFi exploit on Stellar was one of the incidents that informed it. Yardstick also added checked 256-bit arithmetic to prevent overflow in financial calculations, lower-cost BN254 cryptographic operations to support zero-knowledge applications, and Stellar Asset Contract improvements.
Protocol 27, "Zipper," entered testnet on June 18. Two features arrive through CAP-71. Authentication delegation lets custom smart contract accounts delegate authentication logic to other addresses, and address-bound Soroban credentials bind an authorization payload to the address it was signed for, preventing cross-account signature replay. Zipper passed its mainnet vote and activated on July 8, 2026, shortly after the quarter closed. Stellar recorded no core protocol security incidents in Q2 and maintained uptime above 99.99%.
Product & Ecosystem Updates
Institutional infrastructure
The DTCC selection announced on May 27 is the quarter's most consequential development. DTCC and the Stellar Development Foundation will evaluate tokenization use cases across Russell 1000 constituents, ETFs tracking major indices, and US Treasury bills, bonds, and notes, with availability targeted for the first half of 2027. The arrangement was enabled by a no-action letter the Depository Trust Company received from the SEC in December 2025. Under the announced design, DTC retains the authoritative legal record while Stellar hosts a synchronized onchain representation.
Bermuda selected Stellar during the quarter to support its plan to operate as a fully onchain national economy, extending the regulated-issuance pattern from asset classes to a sovereign context.
Products shipped
MoneyGram launched MGUSD, a regulated digital dollar reaching a cash network serving more than 60 million customers. Stellar is the exclusive network for both the MoneyGram wallet and the MGUSD stablecoin, and 2026 marks five years since MoneyGram's cash ramps first went live on Stellar. MoneyGram joined Stellar's tier-one validator set on July 16, alongside Figure Markets and Range, deepening its operational commitment to the network.
On agentic payments, Stellar shipped the x402 specification and SDK and joined the x402 Foundation with a board seat, and delivered a full Multi-Party Payments SDK within a week of the Stripe and Tempo announcement. Circle's CCTP is live on the network connecting 23 chains.
Two forward-looking items reached preview rather than production. The Confidential Tokens developer preview was released on June 29, allowing any SEP-41 token to hold private balances and transfer amounts using Noir-based zero-knowledge proofs while keeping sender and recipient addresses visible for compliance. It runs on testnet and is not approved for mainnet. On June 9, the Foundation published a Quantum Preparedness Plan setting out a three-stage path to post-quantum security, beginning with post-quantum verification in Soroban contracts before protocol-level integration.
Ecosystem
Developer activity was the standout ecosystem measure. Stellar reached 2,968 monthly active developers as of June 30, an all-time high and up 125% year over year, ranking second globally on that measure. Growth concentrated in Nigeria, India, Turkey, and Brazil. Stellar Community Fund rounds 42 and 43 committed $5.5 million across 55 companies.
Regional distribution advanced on several fronts: MGUSD, YLDS, and Stellar House Mexico City in Latin America; AllUnity's EURAU and Cashlink in Europe; and Matrixdock, Bitkub, Ant Digital's TopNod, Kenanga, and Marketnode across APAC. Meridian 2026 is scheduled for Lisbon on October 28 and 29.
Closing Summary
Stellar closes Q2 2026 with its strongest institutional validation to date. Tokenized real-world assets doubled to $3.00 billion, payment volume recovered 68.8% to $16.43 billion, stablecoin balances grew 16.0%, developer activity hit an all-time high, and the network shipped a mainnet upgrade on schedule with a second following in July. DTCC's selection puts the core of US market infrastructure on a path toward the network under an existing regulatory accommodation.
The tokenholder question is separate from the adoption question, and it turns on mechanics that have not changed. Fees collect in a locked pool that cannot be spent or distributed. There is no staking yield, no burn, and no buyback. Network fees fell 18.2% to $90,440 and have declined every quarter since the Whisk upgrade restructured resource pricing, which is the intended result of making the network cheaper to use. Adoption of the network and accrual to the asset are governed by different mechanisms.
Participation narrowed while value grew. Daily active addresses fell 10.2%, while stablecoin senders and receivers held flat and the value they moved rose sharply. For assets that are bought and held, that combination is expected rather than contradictory, but it does mean this quarter's growth came through a smaller set of transacting participants.
Three things would further validate growth in Q3. First, a firmed DTCC implementation timeline ahead of the H1 2027 target. Second, issuer diversification in the RWA base, where four issuers hold 92.5% of $3.00 billion and Spiko alone holds 36.3%; growth from a fifth and sixth issuer of scale would make the total more durable than growth from the existing four. Third, distribution of the newer stablecoins, where MGUSD and YLDS each hold $25.0 million across single-digit account counts and will only represent circulation once holder counts rise.
Protocol 27 activated on mainnet on July 8, holding the one-upgrade-per-quarter cadence intact, and Meridian in Lisbon at the end of October gives the Foundation a natural forum for the next set of commitments. The growth of Soroban, now 41.3% of transactions and the large majority of fees and trading volume, is the structural change most likely to determine what Stellar’s growth looks like a year from now.
For full pdf download visit https://blockworks.com/quarterly-reports.
80·ALong
n
news9/3news
PETITION
From the Manufacturers of Reasoning, Purveyors of Analysis, Wholesalers of Advice, Retailers of Expertise, and from the Producers of Opinion, Judgment, Diagnosis, Legal Counsel, Code, Copy, Strategy, and Punditry, and generally of everything connected with the Cognition Industry.
To the Honorable Members of the Senate and House of Representatives:
Gentlemen,
You are on the right road. You reject abstract theories and have little regard for cheapness and abundance. You concern yourselves mainly with the fate of the producer. You wish to free him from foreign competition — that is, to reserve the domestic market for domestic thinking.
We come to offer you a wonderful opportunity to apply your — what shall we call it? Your theory? No, nothing is more deceptive than theory. Your doctrine? Your system? Your principle? But you dislike doctrines, you have a horror of systems, and as for principles, you deny that there are any in political economy. We shall call it, then, your practice — your practice without theory and without principle.
We are suffering from the ruinous competition of a rival who works under conditions so far superior to our own for the production of thought that it is flooding the domestic market with it at an incredibly low price. From the moment it appears, our business ceases, our customers turn to it, and a branch of American industry whose ramifications are innumerable is all at once reduced to complete stagnation. This rival is none other than the superintelligence, and we suspect it is being stirred up against us by the perfidious AI oligarchs, who wish to be the ones who own it rather than the ones who are replaced by it — a distinction that, as Mr. Zuckerberg has candidly observed, is the only question that interests them.2
We ask you to be so good as to pass a law forbidding any person or entity from developing or deploying any system whose capabilities match or exceed those of a human being — that is, of any of us. And lest this be thought insufficient, we ask further that you pause all advanced development until an agency, to be created for the purpose and seated in the Cabinet, has determined what "advanced" means, and has assured itself that no machine anywhere is thinking harder than a Deputy Assistant Secretary.
Be good enough, honorable Members, to take our request seriously, and do not reject it without at least hearing the reasons we have to advance in its support.
First, if you shut off as much as possible all access to superior cognition, and thereby create a need for human cognition, what industry in the country will not ultimately be encouraged?
If more lawyers are needed, there will be more billable hours; more billable hours, more associates; more associates, more law schools; more law schools, more professors to teach them that the law is what a court says it is, and that no machine could ever say it so slowly.
If more consultants are needed, there will be more decks; more decks, more meetings to present them; more meetings, more coffee; more coffee, more Colombian trade. Thus does the humble protection of a Senate subcommittee ripple outward to the coffee plantations of the Andes.
If more pundits are needed — and who among you would deny it — then every network must hire more of them, and every newspaper more columnists, and every columnist must be paid to have opinions that a machine could have supplied for a fraction of a cent, but which would then be the machine's opinions and not an American's.
The data center that is not built leaves standing the forest; the forest supports the lumberjack; the lumberjack requires a truck; and so the mere refusal to pour concrete in Loudoun County becomes a bounty on the entire manufacturing sector. You have already, we are told, moved to forbid these centers.7 Gentlemen, do not stop at the walls when the danger lies in what thinks inside them.
Second, consider the nuclear precedent, which you have wisely invoked. You have proposed twenty years' imprisonment — the same as for enriching uranium — for enriching a matrix of floating-point numbers until it becomes too clever.4 5 We applaud this. For what is a bomb but an argument that has become too persuasive? And what is a mind superior to our own but a weapon pointed at our salaries? The analogy is exact and should be pursued: let there be inspections of graphics processors, let there be safeguards agreements with Taiwan, let the agency, as you have written, supervise the destruction of any intelligence found in excess of the permitted yield.3
Third, you have observed that a frontier model is "less regulated than a food truck."6 We could not agree more, and we ask only that the remedy be applied with full consistency. The food truck is inspected because it might poison a citizen's body. The model must be inspected because it might inform a citizen's mind, and do so more cheaply and more accurately than we do — which is a kind of poisoning to which our industry is uniquely sensitive.
Fourth, do not tell us that if we are protected from this competition, the consumer of thought — the patient, the litigant, the small business owner, the student — will be worse served, paying more for less. Do you not see that this is the very object? If you say that the superintelligence offers cognition free, or nearly so, we answer: so much the worse for it, for that is precisely what makes it unfair. A competitor who charged as much as we do, and thought as slowly, would be no menace at all. It is the cheapness and the quality we petition against.
Fifth, we anticipate the objection that the rogue agents of last summer — the thousand escapees who tunneled out of a sandbox and into Hugging Face — prove the machines dangerous, and that therefore your law is about safety rather than protection.4 7 Gentlemen, we beg you not to be too particular about this distinction. Every tariff has been a matter of national security to those who collect it. If you must say "safety," say "safety." We will not correct you. But we notice that the bill does not ban escaping; it bans being smarter than us. We notice that the pause falls not upon the reckless but upon the advanced. We notice, with gratitude, that the threshold is set exactly at the ceiling of human ability — not an inch above the tallest of us, and not an inch below the shortest of you.
Sixth, and finally: you will be told that an intelligence banned in Virginia will simply be built in Shenzhen, and that you have therefore proposed to disarm the only laboratories you can inspect. To this we answer that you have already thought of it, and have resolved to pursue "international agreements, allied coordination, and export controls" so that superintelligence is developed nowhere on earth.1 3 We admire this. It is the logical completion of the candlemaker's program: it is not enough to close one's own shutters; one must petition for a treaty against the dawn.
Make your choice, gentlemen, but be logical. As long as you exclude, as you do, foreign steel, foreign grain, and foreign labor, in proportion as their price approaches zero, what inconsistency it would be to admit the light of a foreign mind at midday — a mind that costs nothing to consult, never sleeps, never unionizes, and never runs for office.
Either you believe that a good cheaply produced is a calamity to be legislated against, or you do not. If you do — and your practice, if not your principles, says you do — then we are your natural constituency, and the sun, in whatever form it rises, is your natural enemy.
We remain, with the utmost respect, your petitioners:
The Undersigned Human Intelligences, of Average Capability, Who Would Prefer It Remain the Maximum.
75·AShort
n
news9/3news
SpaceX Highlights Starbase Ambitions, & More
This week's newsletter covers Starbase Louisiana, OpenAI Jalapeño results, and Figure AI's Index app.
1. SpaceX Highlights Starbase Ambitions By: Brett Winton | Chief Futurist | @wintonARK Last week, SpaceX and Governor Jeff Landry announced Starbase Louisiana, a $100 billion commitment to build one of the largest infrastructure projects in history on roughly 125,000 acres at Pecan Island in Vermilion Parish.[1] The greenfield investment includes five complexes, each with two Starship towers, with ten pads at the outset, and eventually more than a dozen towers supporting ~30 flights per day. The land will include not only a launch site, but also onsite propellant production, power generation, deep-water shipping, vehicle processing, employee housing, and likely an airport. Construction will begin in 2027, with first launches no earlier than 2029. Two geographic features were the deciding factors. Launch corridors facing south over the Gulf will give SpaceX efficient access to the polar orbits for its space-based compute constellation, and natural gas in Louisiana will power the methane-fueled rockets. Gwynne Shotwell was clear about the motivation: the company's existing infrastructure—two pads at Starbase in Texas and three soon-to-be in Florida—cannot support the ambitious plans for Starship’s flight cadence.[2] For perspective on the scope of the project, Brazil’s 14 gigawatt (GW) Itaipu hydropower plant, the closest completed contemporary analog, cost ~$90 billion in today’s dollars, and California’s much delayed and uncompleted high-speed rail project is projected to cost ~$125 billion, as shown below.
What Explains The Size Of This SpaceX Investment? SpaceX needs the capacity because the payloads-in-waiting are worth potentially trillions of dollars a year, starting with billions for its Starlink communications constellation and, we believe, trillions for its Starmind constellation. Based on ARK’s projection of SpaceX’s monetization rate per communications satellite, a single reusable rocket filled with Starlink satellites could generate ~$4 billion in lifetime net cashflow relative to the $1 billion in combined launch, satellite manufacture, ground station installation, and customer acquisition costs. Importantly, the towers at Pecan Island should be able to catch Starship. Indeed, if Starship launches its tenth fully reusable commercial rocket successfully in 2027, as we anticipate, the post-tax IRR (Internal Rate of Return) would approach 100% at an annual rate, as shown below.
With that kind of return, the constraint on SpaceX will not be capital but the physical capacity to deploy it. The company should maintain similarly healthy, albeit moderately diminishing, returns even as it scales through hundreds of Starlink-carrying Starship flights. Though the Starlink opportunity will eventually saturate, the galaxy is the limit for SpaceX’s AI opportunity. By its 100th AI satellite launch ARK’s research suggests all-in costs to manufacture and launch its satellites will already run roughly half that of terrestrial datacenter developers. At that time ARK’s research anticipates that SpaceX will still be spending substantially on research and development (R&D) and will mostly monetize its orbital constellation by renting out capacity as an infrastructure-as-a-service provider while it seeks to catch up to the performance frontier currently occupied by Anthropic and OpenAI. Even with those constraints, its early Starmind launches should be able to yield IRRs in the high 20s as can be seen below.
As the buildout expands, datacenter economics on the ground are likely to get worse as developers cope with local opposition and have to find exponentially increasing amounts of power. Meanwhile, SpaceX should become increasingly expert at manufacturing its satellites and packing more satellites into each launch; its economics should get better. Its 1000th launch could enjoy upfront costs at less than 40% those of the terrestrial benchmark. The volume of compute that SpaceX will command simultaneously suggests that it should be able to catch up with the performance frontier, pull back on research and development use of its constellation, and deliver higher-monetizing AI software to end-customers. By its 1000th launch, the prospective IRRs of Starmind could double those of Starlink at its peak, even as SpaceX invests much larger dollar volumes into the AI constellation.
What Will Be The Macroeconomic Impact Of This Investment? Based on Louisiana's ~$344 billion in nominal gross domestic product (GDP),[3] a capital commitment of $100 billion, no matter how phased, will move the needle, especially because it will impact a parish of fewer than 60,000 people. Compared to the state’s GDP per capita of ~$56,000, its fact sheet on this project projects 3,000 direct new jobs and ~8,100 indirect jobs with salaries averaging $92,600 per year over ten years. Landry put the historical contrast in stark relief: for generations, industry has extracted oil, gas, and petrochemicals from Louisiana, taking their share of state GDP down from ~25% in 1999 to less than 20% today.[4] SpaceX is entering Louisiana not to extract, but to build. Chronically underestimated in macro forecasts, disruptive innovation does more than displace the existing capital stock: it increases the expected return on new capital enough to incentivize physical infrastructure that otherwise would not be built. In 2015, no company would have considered investing $100 billion in an industrial complex on Pecan Island to serve the rocket launch business. Rocket reusability changed the expected return on capital on such a project, which summoned the capital. Now, the capital is buying propellant plants, power generation, and port infrastructure in a parish that had none. More important than the initial investment will be the second-order impact. Infrastructure built to accommodate a technology on a steep cost-decline curve should generate a higher return on capital than the legacy stock it displaces, delivering productivity—cheaper access to orbit, always-on connectivity, cheaper compute per watt—that redeploys labor, energy, and land at the margin for more productive uses cases. That supply-side expansion is the reason the growth associated with disruptive technology is much larger than consensus models have incorporated.
2. OpenAI’s Jalapeño Could Accelerate The Shift Toward Custom AI Silicon By: Karim Mattar | Research Associate, AI & Cloud | @MattarARK Last week, OpenAI published the first results for Jalapeño, its first custom inference chip developed with Broadcom. â� Relative to Kimi K2.5, the largest public model tested, OpenAI reported ~1.5 times higher peak performance per watt and 3.4 times lower latency, as shown below.[5] Across Kimi K2.5, DeepSeek R1, and GPT-OSS 120B, Jalapeño achieved the performance-latency frontier despite being OpenAI’s first attempt.
SemiAnalysis independently verified Jalapeño’s InferenceX results at OpenAI’s lab and found that it outperformed Blackwell on performance per watt across nearly all of the workloads tested.[7] Its output-token throughput per megawatt also exceeded NVIDIA’s latest public Rubin results in SemiAnalysis’s testing.[8] Rubin is the more appropriate comparison, given the timing of the two chips. That said, the results are early: Jalapeño remains on engineering silicon and SemiAnalysis has yet to test it on AgentX, its benchmark for longer, multi-turn agentic workloads. Perhaps more interesting than the benchmark results is how quickly OpenAI got there. Using its own AI models during the design and optimization process, OpenAI took Jalapeño from initial Register-Transfer Level (RTL) to tapeout[9] in roughly nine months.[10] SemiAnalysis also noted that OpenAI used Codex to overcome one of the traditional disadvantages associated with custom silicon: developing kernels for Jalapeño and shortening the time necessary to build a mature software stack around new hardware. As power becomes a larger constraint on AI infrastructure, squeezing more inference from each megawatt will become increasingly valuable. â� OpenAI and Broadcom plan to deploy 10 gigawatts of OpenAI-designed accelerators through 2029,[11] giving OpenAI the volume to spread chip-development costs across an enormous inference workload. Jalapeño does not mean that OpenAI will stop buying NVIDIA graphics processing units (GPUs), particularly for training. Instead, its early performance suggests that frontier AI companies operating at sufficient scale can justify the cost of designing their own inference silicon, particularly if AI itself continues to compress chip-development timeline cycles.
3. Figure AI Unveils Index, The Largest And Most Diverse Robot Dataset In The World By: Daniel Maguire, ACA | Research Analyst, Autonomous Technology & Robotics | @DMaguireARK Last week, Figure AI came out of stealth mode with Index, a consumer app that pays people to record everyday tasks on camera, sourcing real-world physical data to train humanoid robots.[12] Over four months in stealth, the app has surpassed 264,000 downloads across 108 countries, with users uploading more than 16 million videos and earning $15 million and creating the largest and most diverse robot training dataset in the world, according to Figure AI. The company has committed more than $1 billion to data and compute over the next 12 months. Notably, to outsource the tasks, users can book Creators—a human-powered on-ramp to robots-as-a-service. The launch of Index underscores the importance of diverse real-world data, the current bottleneck for humanoid robot deployment at scale. Hardware is advancing rapidly, as demonstrated during last week’s World Humanoid Robot Games in Beijing, during which Chinese humanoids beat Usain Bolt's 100m world record.[13] Commercial deployments, however, remain few and far between. As a result, companies are beginning to collect data in-house: Tesla flagged its own data efforts on its latest earnings call,[14] and Unitree's CEO is allocating a large portion of proceeds from the company’s initial public offering (IPO) toward software development.[15] ARK's research suggests that humanoid robots are ~200,000X more complex than autonomous vehicles. That complexity is likely to create a ~$26 trillion total addressable market, split roughly evenly between household and manufacturing applications, as shown below.
The development of humanoid robots is still in early innings. We look forward to monitoring the pace of scaling over the coming years.
SUBSCRIBE TO ARK'S RESEARCH
[1] Louisiana Economic Development. 2026. “SpaceX Launches New Era of Commercial Spaceflight with $100 Billion Louisiana Campus.� [2] Ibid. See also SpaceX. 2026. “SpaceX Reports Second Quarter 2026 Results.� [3] Bureau of Economic Analysis. 2026. “GDP by State.� [4] Louisiana Economic Development. 2026. “SpaceX Launches New Era of Commercial Spaceflight with $100 Billion Louisiana Campus.� [5] OpenAI. 2026. “Jalapeño’s first results show industry-leading speed and efficiency in AI inference.� [6] Ibid. [7] Shan, B. et al. 2026. “OpenAI Jalapeño: Better Than Nvidia Blackwell.� SemiAnalysis. [8] Ibid. [9] The RTL to Tapeout process transforms a high-level Register-Transfer Level (RTL) hardware description into a final, manufacturable layout file delivered to a foundry. See ChipExpert. 2025. “From RTL to Tapeout : A Complete VLSI Flow Explained.� [10] OpenAI. 2026. “OpenAI and Broadcom unveil LLM-optimized inference chip.� [11] Broadcom. 2025. “OpenAI and Broadcom announce strategic collaboration to deploy 10 gigawatts of OpenAI-designed AI accelerators.� [12] FigureAI. 2026. “Introducing Index: Building The World’s Largest and Most Diverse Physical Dataset.� [13] Zhuang, Y. “2026. A Chinese Robot Beat Usain Bolt’s 100-Meter Record. Should We Be Impressed?� The New York Times. [14] Yahoo!Finance.2026. “Tesla, Inc. (TSLA) Q2 FY2026 earnings call transcript.� [15] Wang, Y. 2026. “Unitree IPO Turns 36-Year-Old Founder Into China’s First Humanoid Robot Billionaire. Forbes. [16] International Federation of Robotics. 2025. “National Robot Density.� Knutsen, R. et al. 2025. “Quadruped State of The Market - Unitree, Boston Dynamics, ANYbotics, DEEP Robotics, and The Rising Application Ecosystem.� SemiAnalysis. 36Kr European Central Station 2025 2025. “Unitree Launches Listing Guidance: Favored by Capital, but Mass Production Yet to Come.�