Launch Autopsy #2: Filecoin and the SAFT
The 2017 sale that inverted every choice Ethereum made: accredited-only, lawyer-built, fully vested — a record raise closed three days after China banned ICOs, for a network that was three years away. The SAFT era, in one launch.
Bruno Pierri GalvaoThe first study in this series ended on a warning: every rule Ethereum’s launch proved came with an era attached. This study is about the era that ignored that. In 2017, token sales raised roughly $5.5–6 billion across thousands of projects — re-running Ethereum’s uncapped, unvested, ungated playbook at scale, mostly without Ethereum’s discipline. A widely cited mid-2018 analysis classified some 78% of that year’s as outright scams. The era’s monument isn’t a chain; it’s a casualty list.
Which is why this autopsy picks the era’s most disciplined launch, not its worst. Filecoin — Protocol Labs’ decentralized storage network — ran the sale that inverted every choice Ethereum had made three years earlier: accredited buyers instead of anyone with BTC, everything vested instead of nothing, lawyers before code, and a new legal instrument — the — invented specifically for it. It became the largest raise of the boom’s peak, and the cleanest test of the question this series keeps asking: how much does the mechanism matter, once the era has had its say?
The world it launched into
Ethereum sold into a bear market and open hostility. Filecoin sold into the opposite — a mania with a regulator finally in the room.
Two weeks before the sale opened, the SEC published the (July 25, 2017): tokens can be securities, judged case by case. The free-for-all Ethereum had slipped through was closing. Meanwhile the boom Ethereum’s own success had ignited was setting records monthly — raised ~$153M in June, ~$232M in July. ETH traded near $300, Bitcoin was sprinting toward $5,000, and the parabola that would end near $19,700 in December was already forming.
Then, mid-sale, the other shoe: on September 4, 2017, seven Chinese regulators led by the declared ICOs illegal fundraising and ordered refunds. Bitcoin fell roughly a quarter in ten days. Filecoin’s sale closed three days later — a record raise, completed inside the week the first great crackdown began.
That’s the era in one frame: maximum greed and the first real fear, overlapping by 72 hours. Filecoin’s entire design — the SAFT, the accreditation gate, the vesting — was built for exactly this world: it assumed the DAO Report meant what it said, and it treated legal survival as a launch requirement rather than an afterthought. Most of the 2017 class made the opposite bet. Very few of them still exist.
The mechanics
The advisor round came first: ~$52M from roughly 150 investors — Sequoia, Union Square Ventures, a16z, Winklevoss Capital, DCG, Y Combinator among the names on the record — at a flat $0.75 per FIL, with mandatory vesting of one to three years and discounts of up to 30% scaling with the lockup.
The public sale ran August 10 to September 7, 2017, on — a platform that did not exist a year earlier, spun out of AngelList jointly with Protocol Labs specifically to run this sale compliantly. Participation required and under 506(c); what buyers signed was not a token purchase but a — a security promising future delivery of tokens at network launch, resale-restricted for a year.
Design choices worth naming plainly:
- Price as a function, not a number. The sale price rose mechanically with
the total raised:
price = max($1, raised / $40M)per FIL. Demand didn’t just fill the sale, it repriced it — from ~$1.30 when the public window opened (the advisor round already counted) to roughly $5 at the close. And it repriced in real time: each confirmed payment raised the price for every payment behind it in the queue, while some cryptocurrencies took an hour or two to confirm. The crowd set its own terms; latecomers paid ~4× what hour-one buyers did. - Vesting everywhere. Public buyers chose their own lockup, and the sale priced it: 6 months of linear vesting bought no discount, 1 year bought 7.5%, 2 years 15%, 3 years 20%. Advisors vested one to three years; Protocol Labs and the Filecoin Foundation vest over six. Nothing — literally no allocation — was liquid at network launch. Ethereum’s sale had zero vesting anywhere; Filecoin’s had it everywhere, with the discount curve putting an explicit market price on patience.
- A sliver of supply sold. The 2017 sale was capped at 200M FIL and ended up accounting for 7.5% of the 2B max supply on the launched network’s books. Ethereum sold 83.5% of genesis to the public; Filecoin kept 70% for decades of future mining rewards, released on a six-year half-life — Bitcoin’s schedule, stretched. This single inversion — sell almost everything vs. sell almost nothing — is the era’s biggest structural shift, and it decided the shape of everything that happened after 2020.
- The gate as a feature. 2,100+ accredited investors is not a distribution; it’s a cap table. The exclusion wasn’t a side effect of compliance — it was the compliance. The SAFT’s whole theory was that the public never touches the security; the public’s absence was load-bearing. One detail makes the point perfectly: the minimum investment was $10. The gate was never about the size of your check — only about the regulator-defined size of your net worth.
- Demand crushed the infrastructure anyway. $135M arrived in the first hour of the public window and CoinList buckled; the sale paused for more than a day to verify transactions before resuming. Even a sale built to slow money down couldn’t slow 2017 money down.
The raise itself needs an honest footnote. The number that made headlines — “$257M, the largest ICO ever” — marked the incoming crypto at its September peak. Protocol Labs’ own accounting, valuing assets as they were invested, was $205.8M including the advisor round — which is less than Tezos’ $232M. In 2017, even the size of your raise was a function of the market’s mood on settlement day. The era’s record-keeping was as volatile as its assets.
The whole package on one card — same anatomy as every study in the series:
Note what the sinks column says: unlike Ethereum — one sink at launch, the rest arriving over seven years — Filecoin shipped with real demand sinks on day one. Storage providers must lock to mine at all; is burned -style; storage deals are paid in FIL. On paper, the strongest launch-day demand story of any token in this series so far. Hold that thought.
What happened next
First, the wait. At sale time the network was targeted for 2018–2019; testnet arrived December 2019, and mainnet on October 15, 2020 — launched at block 148,888, a nod to the Chinese mining community that dominated its early capacity. Three years and five weeks separated the money from the network. The SAFT made that gap legal. Nothing could make it free: buyers carried schedule risk the whole way down, locked in a security they couldn’t resell for a year and holding claims on a token that didn’t exist.
FIL sale price · vs cumulative raised · Aug 10 – Sep 7, 2017
Then, the launch — and the launch-day irony this study turns on. Filecoin’s proudest design feature, the day-one demand sink, immediately attacked its own supply side. Miners needed FIL as before they could mine, but rewards vest over months and almost nothing was circulating — the network’s operators had to buy, on the open market, the token they were promised for operating it. Within days of mainnet, several of the largest miners powered down in protest (Protocol Labs disputed the word “strike”); the protocol answered by making 25% of block rewards liquid immediately. The strongest sink in the series had worked exactly as designed — against the network itself.
Then, the price. FIL’s opening day traded anywhere from $37 to $100 depending on the venue before settling near $40 — on a float of well under half a percent of eventual supply; it wouldn’t cross even 1% until the end of the month. Five and a half months later, on April 1, 2021, it printed ~$237: an implied around $474 billion — briefly pricing a storage network with negligible paying demand above almost every company on Earth. This is doing what low float does: price discovery on a sliver, with a decade of unlocks and mining emissions scheduled to sell into whatever number the sliver hallucinated.
FIL/USD · mainnet to the bear low
What followed the ATH was arithmetic. Emissions and unlocks ground the price down for two years; the December 2022 bear-market low, around $2.60, landed almost exactly on the public sale’s estimated average price — roughly $2.6 per FIL. (Protocol Labs never published a token count or an average; that figure is derived from the sale’s price curve. Blend in the $0.75 advisor round and the average across the whole raise drops to about $1.40.) Note the structure of that near-miss: for the entire vesting window — every tranche, 6 months to 3 years, fully unlocked by late 2023 — spot stayed at or above the estimated public-sale average. A disciplined buyer could have exited at or above cost at any point while their tokens unlocked. The decisive break below the sale price came after vesting ended — and kept going. As of mid-2026 FIL trades around $0.73: below the sale’s $1 floor, below even the advisors’ $0.75, down ~99.7% from the top. Ethereum’s launch never put a single ICO buyer underwater, even at its bottom; Filecoin’s, nine years on, has put all of them there.
One more ending belongs to this study, though it happened to someone else. The SAFT’s legal theory — the agreement is a security, the delivered token is not — was never blessed so much as tolerated. In 2020, a federal court blocked ’s $1.7B SAFT-funded token distribution, ruling the whole scheme one securities offering; Telegram returned $1.2B and abandoned the project. Filecoin itself was never sued — the instrument did its narrow job for the project it was invented for — but the door it opened closed behind it. The era’s signature innovation lasted one era.
What worked
- It shipped, and it still runs. Against a 2017 cohort where the canonical study found ~78% scams, Filecoin delivered a working network with real cryptography — replacing wasted hashing — and it operates at scale today. The bar was on the floor; Filecoin cleared it by a lot. Credit where due.
- Legal survival, by design. Tezos spent its first year in governance war and securities litigation (settled for $25M); lost to the SEC; Telegram was unwound. Filecoin — the biggest target of the class — was never sued. The SAFT+Reg D architecture was built for exactly one storm and it held in exactly that storm.
- Sinks at genesis. The collateral-gas-payments triad meant the token had a job on day one. The implementation drew blood (see below), but the principle — launch with the demand story live, not promised — is the right correction to Ethereum’s gas-only genesis.
- Vesting discipline held. No insider dump at listing was possible because no insider was liquid at listing. The six-year Protocol Labs and Foundation schedules were, at the time, the longest serious lockups in the industry — and they were honored.
- The sale mechanism was honest about greed. The rising-price function meant the sale couldn’t be “won” by insiders front-running a fixed cheap price; demand repriced everyone in real time, in public.
What broke, or nearly did
- Three years of schedule risk, transferred to buyers. The sale assumed delivery in 2018–2019; the chain came in late 2020. Locked, resale-restricted buyers carried that slippage with no exit. The SAFT priced legal risk carefully and delivery risk not at all — and delivery was the risk that materialized.
- The sink that ate its miners. Requiring operators to source the token from a market that barely existed inverted the launch: the people the network needed most were its most desperate forced buyers. A demand sink that gates your own supply side is a tourniquet — it holds price by strangling the network. The day-one patch (25% liquid rewards) was an admission that the economics shipped broken.
- Low float manufactured the wreckage. Selling 7.5% and vesting everything was the responsible choice by 2017’s lights — and it mechanically produced the $237 absurdity, the half-trillion FDV, and the 95%+ drawdown that followed. This is the study’s uncomfortable core: the discipline meant to protect the market is what armed it. Ethereum’s reckless full float never allowed a fantasy price; Filecoin’s careful sliver guaranteed one.
- A cap table is not a community. 2,100 accredited investors bought legal safety at the price of the wide believer base Ethereum’s open sale minted. When the grind came, Filecoin had holders with cost bases and exit discipline, not evangelists — the loyalty dividend this series identified in study #1 was structurally excluded here.
- The demand story is still pending. Collateral and gas create token demand only insofar as someone pays for storage; paid organic demand stayed thin for years (much early usage rode subsidized verified-deal programs). Sinks amplify demand; they cannot replace it. The price today is the market’s estimate of that difference.
Scorecard
| Dimension | Grade | Note |
|---|---|---|
| Distribution fairness | D | 2,100 accredited buyers; exclusion was the design |
| Insider overhang | B+ | Everything vested, insiders 6yr — but 92.5% of supply non-public |
| Demand sinks at launch | B | Real triad on day one; implementation broke its own miners |
| Demand sinks, eventual | C | Sinks without underlying paid demand; still pending |
| Emission schedule | B− | Decades-long, partly growth-contingent; low float side effect |
| Treasury management | Incomplete | Never disclosed — itself a datum for a $205.8M raise |
| Regulatory posture | A− | Built for the storm, survived it; its theory died with Telegram |
| Macro timing | A/C | Raise top-ticked the mania; token listed into the next one — and its ruins |
The series hypothesis — macro decides the outcome, mechanism decides survival — takes a point from each column here. The era gave Filecoin a record raise and a fantasy ATH the design didn’t earn; the mechanism (vesting, sinks, legal architecture) is why the project outlived the class of 2017. And one new corollary: mechanism can also decide the shape of the price path, not just survival — Filecoin’s chart is its float schedule, drawn in market prices.