Within a single day, the digital asset market got two answers it had been waiting months for.

00 — Editor's Letter

Within a single day, the digital asset market got two answers it had been waiting months for. The first came from the US Senate: the market structure bill was rejected, 49 votes for and 50 against. The second came from the bond market: the 10-year yield touched 5.00%. A regulatory door closed, and the cost of capital for every asset in the world got more expensive once again.

On the vote, the number worth remembering is not 49 or 50. It is zero: not one Democratic or independent senator voted in favour, including those who had spent a year at the table negotiating that very text. The reason they gave had nothing to do with crypto and everything to do with a public-ethics provision. Section 04 has the full tally and what room, if any, is left.

On interest rates, the tidiest number sits at the two-year maturity — the maturity that tracks Fed policy most closely. It is at 4.67%, which is 92 basis points above the Fed's own upper bound. The market is not forecasting that the Fed will hike. The market has already priced in that the Fed has to hike, and more than once. The Fed announces its decision at 01:00 tomorrow morning Vietnam time.

The proximate cause is the oil price. Brent crude rose 22.50% over fifteen sessions, pushing the US consumer price index for August up +0.4% month on month — more than a third of that increase came from petrol alone. But there is a detail few people noticed: core inflation year on year actually fell, from 2.5% to 2.4%. Two numbers telling two opposite stories, and both are correct. Sections 01 and 02 pull them apart.

The rest of the picture: twelve of the twenty cells in the indicator table are red. Gold lost 7.69% and silver 6.52% in fifteen sessions alone — when real rates rise, metals that pay no yield get sold first. Australian equities lost 3.88% and Indian 3.52%. In the middle of that red, the VN-Index rose 1.10% to 1,811.15 points, one of eight green cells — there is a specific reason for that, and section 01 gives it.

A note on timing so that you read the numbers in this issue correctly: this newsletter closes its data at the end of the 15/09 session, that is, immediately after the vote and immediately before the Fed meeting. Most of Bitcoin's fall on 15 September happened before the Senate voted — section 04 sets out that sequence clearly, because getting the causality backwards is the fastest way to draw the wrong lesson.

There are four new things in this issue. Section 05 sets out four six-to-twelve-month scenarios from Fable, our Chief Strategy Officer, with an "FOMC switch" box so you can read tomorrow morning's news for yourself. Section 06 profiles Ethereum — and finds a paradox: the network has more users than it has ever had, and earns less money than it has ever earned. Section 07 asks a question this industry rarely asks head-on: what the research actually says about putting part of a portfolio into crypto, and where that research is weak. Section 08 is an interview with the founder of Cosmos Digital, Alex Pham, about five years in digital assets — including where he got it wrong.

---

01 — Global Equities & Macro

Twenty indicators, measured over fifteen sessions, as of the close on 15/09/2026.

• S&P 500 — 7,586 · −1.19%
• Euro Stoxx 50 — 6,236 · −3.39%
• FTSE 100 — 10,658 · −2.10%
• Nikkei 225 — 63,493 · −3.11%
• Hang Seng — 24,918 · −2.35%
• Shanghai — 3,885 · +0.09%
• Kospi — 6,684 · −0.19%
• Taiwan — 45,863 · +2.46%
• Singapore — 5,718 · +0.66%
• India · Nifty 50 — 23,398 · −3.52%
• Australia · ASX 200 — 8,750 · −3.88%
• Vietnam · VN-Index — 1,811.15 · +1.10%
• Gold — $4,334 · −7.69%
• Silver — $64.16 · −6.52%
• Brent crude — $108.51 · +22.50%
• Dollar · broad basket — 118.21 · +0.13%
• 10-year yield — 5.00% · +36 bp
• VIX · fear — 17.20 · +11.33%
• USD/VND — 25,958 · −0.58%
• Crypto market cap — $2.582 trillion · −3.92% (BTC)

Measured against the 15-session mark, as of the close on 15/09/2026. Sources: Yahoo Finance (indices, commodities, VIX, exchange rates) · FRED (broad dollar, 10-year yield) · DNSE/VNDirect (VN-Index) · CoinMarketCap (crypto market cap). The small line in each cell is the last 30 sessions; the horizontal dashed line is the level at the 15-session mark. The crypto market cap cell records the percentage for Bitcoin over the same window, because total market capitalisation has no consistent daily series.

Twelve red cells, eight green, and one up twenty-two per cent. This table does not describe a market that is frightened — it describes a market repricing the cost of money.

Fed · Nine Months on Hold, and a Market That Is Not

The Fed cut rates three times in 2025 — on 18/09, 30/10 and 11/12 — 25 basis points each time, bringing the upper bound from 4.50% down to 3.75%. It has changed nothing since. But over those nine months on hold, the bond market has moved steadily in the opposite direction.

The three Fed rate cuts are marked with dashed lines. The horizontal dashed line is the Fed upper bound; the two solid lines are the 2-year and 10-year US Treasury yields. Sources: yields from the US Treasury's official interest rate tables, as of 15/09/2026; Fed upper bound from FRED.

Three numbers worth writing down:

• 2-year: 4.67%. That is 92 basis points above the Fed upper bound. This is the maturity that tracks Fed policy most closely, so the gap is precisely what the market is saying: rates have to go up.
• 10-year: 5.00%. A twelve-month high, set in today's session. The low of the same series was 3.97% on 27/02/2026 — so in six and a half months it has climbed almost a hundred basis points while the Fed has not moved at all.
• 30-year: 5.36%. And the 10-year minus 2-year spread is +33 basis points — the curve has finished un-inverting and has steepened again, the shape you usually see when the market is pricing persistent inflation rather than recession.

A central bank can set the overnight rate. It cannot set the ten-year rate. When those two move in opposite directions for nine months straight, the one that usually gives way is the short end.

Global Equities · Red, But Not Panicked

The S&P 500 lost 1.19%. But outside the United States it was redder still: Euro Stoxx 50 −3.39%, FTSE 100 −2.10%, Nikkei 225 −3.11%. What stands out is that the two biggest losers were two large domestic markets: Australia −3.88% and India −3.52%. The VIX sat at 17.20, up 11.33% but still in normal territory — this is not a panic, it is an orderly de-rating as the cost of capital goes up.

Asia · Split in Two

Hang Seng −2.35% and Kospi −0.19%, but Shanghai edged up (+0.09%) and Taiwan rose 2.46% — the only two green cells in the equity group. Taiwan is where the semiconductor supply chain is concentrated; when the rest of Asia is sold on interest rates, money stays where the real order book is.

Vietnam · A Green Cell Ahead of 21/09

The VN-Index rose 1.10% to 1,811.15 points, against the entire region. And USD/VND fell 0.58% to 25,958 — the dong strengthened while the broad dollar was essentially unchanged (+0.13%). The date to keep in mind: on 21/09, FTSE Russell's upgrade of Vietnam to secondary emerging market status takes effect. Index-tracking funds buy the index, not their own assessment of it — that is mechanical flow, and markets usually run ahead of the effective date. We will measure the actual foreign flows after 21/09 and report back in issue A04.

Gold · Silver · Oil — Three Cells That Tell the Whole Rate Story

Gold −7.69% to $4,334 and silver −6.52% to $64.16, while Brent crude rose 22.50% to $108.51. Read together, these three cells explain everything else: dearer oil pushes headline inflation up, headline inflation pushes yields up, yields push real rates up — and precious metals, which pay their holders nothing, get sold first. These are not three separate events. This is one causal chain, read from right to left.

Crypto · Bridge to Section 03

Bitcoin lost 3.92% over the same fifteen sessions — lighter than gold, and lighter than Indian or Australian equities. Widen the window to thirty days, though, and the whole crypto group is deep green: Bitcoin +20.44%, Ethereum +27.99%, Solana +29.92%. Two time windows, two opposite answers, and that is exactly why this newsletter always states the date it is measuring from.

---

02 — World Finance (Macro)

August inflation rose because of petrol. Core inflation fell. Both sentences are true, and any reader who hears only one of them will bet the wrong way.

US August Inflation · Released 11/09

US inflation for August 2026, released 11/09. Headline CPI +0.4% month on month and 3.4% year on year; core CPI 2.4% year on year, down from 2.5%. Petrol rose 3.9%, accounting for more than a third of the headline increase. Source: US Bureau of Labor Statistics.

Core is the measure with food and energy stripped out — the one the Fed looks at when it decides, because it reflects persistent price pressure rather than a supply shock. Core is falling. Headline is rising, and it is rising because Brent crude went up 22.50% in fifteen sessions. Do not write, and do not read, the bare sentence "US inflation is rising" — it is half right, and the other half is the half the Fed votes on.

A common confusion. CPI and PCE are two different measures, built from two different baskets, and the Fed's 2% target is set on core PCE, not CPI. The August PCE figure is released on 30/09 — after this issue.

Rates Box · As of 15/09

• Fed upper bound — 3.75% · held since 11/12/2025
• 2-year yield — 4.67% · 92 bp above the Fed upper bound
• 10-year yield — 5.00% · twelve-month high
• 30-year yield — 5.36% · —
• 10-year minus 2-year — +33 bp · the curve has steepened again
• Dollar · broad basket — 118.21 · +0.13% · 15 sessions

Macro Indicator Table

Eleven US macro indicators: the 10-year yield, the 2-year yield, the spread between the two maturities, the effective federal funds rate, the Fed balance sheet, reverse repo, the Treasury General Account, M2 money supply, the dollar index, Brent crude and gold — each row with its own as-of date. Sources: FRED and Yahoo Finance.

Each row carries its own as-of date, because each series is published on a different rhythm — the Fed balance sheet and the Treasury account weekly, money supply monthly, interest rates daily. Collapsing them all onto a single date is the easiest way to say something false. · FRED · Yahoo Finance (gold, oil)

Strategy · The World's Largest Bitcoin Treasury, and Who Gets Paid First

Strategy holds 845,050 bitcoin at an average cost of $75,412 each. At today's price that stack is worth $64.12 billion; add $6.54 billion in cash and you get $70.66 billion in liquid assets — against a market capitalisation of $54.87 billion. On the face of it, the market is valuing the company below the assets it holds. But that calculation misses a layer: preferred stock.

• STRC · Nasdaq · 104,894,705 shares · liquidation preference $10.49 billion
• STRK · Nasdaq · 14,020,744 shares · $1.40 billion
• STRD · Nasdaq · 14,024,221 shares · $1.40 billion
• STRF · Nasdaq · 12,839,689 shares · $1.28 billion
• STRE · Luxembourg · 7,750,000 shares · $0.88 billion
• Total · 153,529,359 shares · $15.46 billion

Five tickers, total liquidation preference $15.46 billion as at 30/06/2026 — nearly double the $8.03 billion at the end of 2025. These are claims that rank ahead of common shareholders: if the company were wound up, they get paid first. Subtract them and $55.20 billion is left for common shareholders, against a market capitalisation of $54.87 billion. The real discount is about 2%, not the twenty-odd per cent you get from a calculation that ignores the preferred layer. Note that this par value floats — it is the greater of the trading price and $100 — and the $15.46 billion figure is as at 30/06, to be updated when the third-quarter report lands in early November. · Form 10-Q, Q2 2026, SEC

In the week ended 07/09, the company neither bought nor sold a single bitcoin, and sold no shares through its at-the-market programme either. It spent $176.3 million in cash buying back 1,810,885 STRC preferred shares, and the board doubled the buyback authorisation from $1 billion to $2 billion. The week before, it had funded the same activity from share sales; this week it came out of the till. The entire bitcoin stack is currently sitting on an unrealised gain of roughly $0.39 billion, or 0.6% — far thinner than the usual picture of this company. · Form 8-K, 08/09/2026, SEC

Metaplanet · When Shareholders Push Back, Management Retreats Twice

Japan's bitcoin treasury company has a management option pool, issued in February 2023 when the company was still called Red Planet Japan, with revenue of ¥366 million and an operating loss of ¥858 million. The original terms: 46 million shares. But that pool had an adjustment mechanism anchored to 20% of fully diluted shares outstanding — so every time the company issued more stock to buy bitcoin, the incentive pool inflated automatically alongside it. Shares outstanding went from 114.7 million to 1.28 billion in three and a half years; the pool rose with them to 319.46 million shares.

• Original terms, 02/2023 — 46,000,000 potential shares · —
• Under the 20% mechanism, at 30/06/2026 — 319,464,000 · ×6.9
• 18/08/2026 — the 20% mechanism is abolished and the number fixed — 319,464,000 · frozen
• 11/09/2026 — cut further — 188,190,000 · −41.1%
• — of which the unexercised portion — 105,366,000 · −55.5%

The notable thing is not that the mechanism existed — it may well have been fully disclosed from the start. The notable thing is that the company retreated twice in three weeks, and also scrapped a plan to move up to 90,000 options into a long-term employee incentive programme. Set alongside Strategy, two companies in two countries with two completely different mechanisms still tell the same story: a treasury machine issues shares to buy assets, and value flows to a group that is not the common shareholder — at Strategy it is the preferred holders collecting a 12% annual dividend, at Metaplanet it is the management incentive pool. The difference is that at Metaplanet, that flow has just been partly blocked. · Metaplanet disclosure, 11/09/2026, TDnet

---

03 — Crypto Market Overview

• Bitcoin — $75,879 · −3.92% · 15 sessions
• Ethereum — $2,406 · −3.07% · 15 sessions
• Solana — $97.40 · −1.48% · 15 sessions
• Total market capitalisation — $2.582 trillion · Bitcoin accounts for 59.02%
• Ethereum share — 11.37% · everything else: 29.60% of the total market
• Fear & Greed — 69 · Greed

Every figure is measured over 15 sessions — exactly the gap between two issues of this newsletter. Sources: CoinGecko · CoinMarketCap · Alternative.me, as of 15/09/2026.

Over the last fifteen sessions, Bitcoin fell less than gold and less than most equity indices outside the US. Widen the view to twelve months and the picture is entirely different.

Bitcoin · Twelve Months

Daily closing prices, twelve months. Peak of $124,740 on 07/10/2025, trough of $58,566 on 01/07/2026. Source: CoinGecko.

Peak to trough is −53%. Trough to today is +31%. And peak to today is still −38.4%. Those three numbers all describe one asset, within the same year. Someone who bought last October and someone who bought this July are living in two different realities — while holding the same thing. Section 07 returns to exactly this point.

Bitcoin ETF Flows

• 08/09 — −$46.6 million
• 09/09 — −$120.2 million
• 10/09 — −$282.7 million · heaviest outflow session
• 11/09 — −$13.2 million
• Four sessions combined — −$462.7 million
• 14/09 — +$159.9 million · IBIT +134.3 · FBTC +53.3 — all remaining funds negative

The four-session outflow streak broke on 14/09. But the arithmetic says something more: IBIT and FBTC together came to +$187.6 million, more than the industry total — meaning the remaining funds were still being redeemed. Money has not returned to the market; it is concentrating in the two largest names. · Farside Investors

Source: Farside Investors, US spot Bitcoin ETF flow table. Figures in brackets are negative. The 15/09 session was not yet available at the time this issue closed.

Derivatives · Funding

The Bitcoin perpetual funding rate is at 0.0099% per 8 hours, roughly 10.80% annualised. A positive number means those betting on higher prices are paying those betting on lower prices — the normal state of a market that is not panicking. This level shows neither leverage being blown out nor a market running for the exit. · Binance

Top Gainers and Losers · 14 Days

Top 10 gainers · top 250 by market cap

• STONK +805.7%
• AKE · Akedo +200.6%
• 龙虾 · 龙虾 (Lobster) +175.4%
• USELESS · Useless Coin +87.7%
• PRL · Pearl +74.8%
• BTW · Bitway +62.4%
• GRX · GRX Chain +62.1%
• EDGE · edgeX +58.5%
• 牛来 · 牛来 (Niu Lai) +52.3%
• PONS +50.6%

Top 10 losers · pegged assets filtered out

• CARDS · Collector Crypt −32.7%
• JASMY · JasmyCoin −25.5%
• CASHCAT · Cash Cat −25.4%
• CVX · Convex Finance −24.3%
• SKR · Seeker −23.0%
• FARTCOIN −19.4%
• PENGU · Pudgy Penguins −19.2%
• CC · Canton −19.1%
• PUMP · Pump.fun −18.8%
• SPX · SPX6900 −18.2%

14-day moves within the 250 largest coins by market capitalisation; 249 had complete data. The losers list has pegged assets filtered out. Source: CoinGecko, as of 15/09/2026.

Market Breadth

Bitcoin accounts for 59.02% of total market capitalisation, Ethereum 11.37%, and everything else combined only 29.60%. Over fifteen sessions all three majors were red: Bitcoin −3.92%, Ethereum −3.07%, Solana −1.48% — Solana fell the least, but over a full year the order reverses completely: Bitcoin −34.2%, Ethereum −46.8%, Solana −58.4%. Fifteen sessions is not an altcoin season; it is just fifteen sessions.

One index that needs its timestamp read alongside it. The fear and greed index is at 69 — greed territory. But this index closes once a day at 00:00 UTC, that is, before the whole of the 15/09 session — before the CLARITY vote, before the 10-year yield touched 5.00%, and before Bitcoin's fall of more than 4%. The number 69 describes yesterday's sentiment, not the sentiment at the moment you are reading. The updated figure appears tomorrow morning.

---

04 — Crypto Movements (Protocols & Exchanges)

● US Senate Rejects the CLARITY Act · 49–50

At 14:19 on 15/09 US Eastern time, the Senate held a procedural vote (cloture) to open consideration of the Digital Asset Market Clarity Act. That procedure requires 60 votes in favour. The result: 49 for, 50 against, 1 not voting. The bill was not merely short of 60 — it failed to reach a simple majority.

• For — 49 · all Republicans
• Against — 50 · 44 Democrats + 2 independents + 4 Republicans
• Not voting — 1 · Chris Coons (Democrat · Delaware)
• The four Republicans voting against — Susan Collins (Maine) · Josh Hawley (Missouri) · Jerry Moran (Kansas) · Thom Tillis (North Carolina)
• Roll call 234 · 119th Senate

The number that matters is not 49 or 50. It is zero: not a single Democratic or independent senator voted in favour. Even those who had spent a year negotiating this text — Kirsten Gillibrand, Mark Warner, Cory Booker, Ruben Gallego — voted against. This is not a bill a few votes short. This is a bill that lost an entire side.

The reason they gave has nothing to do with market structure and everything to do with public ethics: Democrats wanted a provision barring sitting officials and their families from profiting from digital assets. Senator Mark Warner: "We cannot pass foundational legislation for this industry while the President of the United States is permitted to personally profit from it." On the other side, Senator Cynthia Lummis, speaking immediately before the vote: "Let this not be the day we hand our future to someone else simply because we were too afraid to finish what we started."

Is there any door left? Yes, but a narrow one. Thom Tillis voted against on procedure in order to retain the right to move to reconsider — only a senator on the prevailing side has that right — and filed the motion at 15:01 the same day (per The Defiant; we have not been able to verify this against the floor record). That means a second cloture vote could happen without restarting the procedure from scratch. But the calendar does not help: the House has cancelled its work weeks of 21/09 and 28/09, the Senate is in a state work period from 05/10 to 06/11 and only reconvenes on 11/11, after the 03/11 midterm elections. That leaves fewer than three session weeks before this year's window closes.

Prediction markets reacted almost instantly: the probability of CLARITY becoming law in 2026 on Polymarket fell from 29.5% on Monday to 6.5%; Kalshi had it at 8%. The House passed the original H.R. 3633 by 294–134 back in July 2025 — more than fourteen months ago. The gap between the two chambers, rather than the gap between the two parties on crypto, is the one that has not been closed.

How to read today's price chart correctly. Bitcoin fell sharply on 15/09, but most of the fall happened before the vote. The overnight high was $79,530; by 10:50 a.m. — more than three hours before the vote — it was already down to $75,750. Two other forces weighed on the same day: the 10-year Treasury yield touching 5.00%, and the Fed meeting the very next day. Saying "CLARITY failed and Bitcoin crashed" is neat but has the sequence wrong. More accurate: a market already stretched by interest rates, meeting a regulatory door closing.

· Roll call vote 234, United States Senate· CoinDesk · The Defiant

● Connecticut Orders Nine Prediction Platforms to Halt Sports Contracts

On 10/09, the Connecticut Department of Consumer Protection issued immediate cease-and-desist orders against nine platforms: Polymarket, Coinbase, Crypto.com, Robinhood, Gemini, Webull, ProphetX, Novig and Underdog Predict, alleging they offered unlawful sports event contracts. The state simultaneously issued subpoenas to nine licensed service providers, including PayPal and Plaid. Kalshi was not in the cease-and-desist group — the state has sued that company in a separate action.

This is the first time a US state has swept both a major crypto exchange and a listed brokerage into the same gambling order. · Connecticut Governor's Office statement

● Circle Acquires Tazapay

On 08/09, Circle signed a definitive agreement to acquire Tazapay — a Singapore cross-border payments company handling more than $25 billion in annualised volume, of which roughly 60% already runs on stablecoins, with more than 60 banking and fintech partners across more than 100 payout markets. The deal is expected to close during 2026 and requires approval from the Monetary Authority of Singapore. Circle did not disclose the value of the transaction.

The world's second-largest stablecoin issuer is buying payout pipes rather than waiting for others to connect to it. · Circle statement, SEC filing

● Trezor: Number of Customers Exposed Rises to 80,689

On 04/09, Trezor raised the number of customers whose data was exposed through logistics contractor ShipMonk to 80,689 — roughly 67,000 more than the 13/08 disclosure, after discovering that the contractor had still been storing 2019–2021 order records despite having confirmed deletion in writing. For 11,742 customers, the exposed data included full name, shipping address, phone number, email and order number. Devices, private keys and recovery phrases were unaffected.

The hardware wallets are still safe. But the list of people who bought hardware wallets sits with a third party, and that list is a perfect target for impersonation scams — a link in the chain no user controls. · Trezor notice

● Bitwise Closes Its Dogecoin ETF

On 10/09, Bitwise announced the liquidation of the Bitwise Dogecoin ETF (NYSE Arca: BWOW), less than a year after listing. The last trading day is 14/10/2026; shareholders receive cash at net asset value struck on 21/10, paid on 22/10. Bitwise gave product line optimisation as the reason.

The wave of single-asset digital asset ETFs opened very quickly over the past two years; this is one of the first to close. Whether a financial product survives depends on whether enough people pay the management fee, not on how famous the underlying asset is. · Bitwise statement

---

05 — Market Trends (Four 6–12 Month Scenarios)

Data as of the close on 15/09/2026, BEFORE the FOMC announcement at 01:00 on the morning of 17/09 Vietnam time. Weightings will be updated in issue A04.

The four scenarios below are a six-to-twelve-month story, out to the third quarter of 2027. The Fed's decision tomorrow morning does not change the name of any scenario — it only changes the weightings. So we are printing all four at once, with a switch you can use to adjust them yourself once you know the outcome.

This issue's central axis fits in one sentence: does the Fed win, or does the bond market win. The Fed says inflation is coming home — core year on year has fallen from 2.5% to 2.4%. The bond market says not yet, and demands a two-year rate 92 basis points above the Fed's upper bound. One of the two will have to give ground.

Four Scenarios · Fable, Chief Strategy Officer

• S1 · Delayed soft landing — 30%. The Fed hikes 25 basis points on 16/09 and then stops for good. Core PCE inflation turns in the fourth quarter. The 10-year yield falls back below 4.5% in the first quarter of 2027. Bitcoin clears the $83,000–86,000 upper band in Q1–Q2 2027; altcoins recover selectively.
• S2 · Higher for longer (base case) — 35%. A 25 basis point hike, then on hold at 4.00–4.25% until mid-2027. The 10-year yield is stuck in a 4.6–5.0% range. Bitcoin trades between $62,000 and $86,000; ETF flows swing on headlines; Bitcoin dominance stays high, altcoins stay weak.
• S3 · Second wave — 20%. Oil and producer prices push headline inflation through 4% and household expectations come unanchored. The Fed hikes 50 basis points in total by December. The 10-year yield goes through 5.2%. Bitcoin loses the $62,000 floor and tests the $52,000–58,000 zone; recovery comes late, at the end of 2027.
• S4 · Something breaks on leverage — 15%. Rates stay high long enough to break a link in the debt chain — AI capital expenditure, private credit, or a crypto treasury company. Junk bond spreads blow out and the Fed cuts on an emergency basis. Bitcoin falls around 30% first, then recovers in a V-shape on liquidity.

✦ FOMC switch — use after 01:00 on the morning of 17/09

If the Fed HIKES 25 basis points (the market is pricing roughly 80%): S1 30 · S2 35 · S3 20 · S4 15

If the Fed HOLDS: S1 40 · S2 25 · S3 20 · S4 15

A hold makes S1 heavier because it would mean the Fed believes core inflation is coming home. The two tail scenarios, S3 and S4, do not change — they depend on the oil price and the structure of debt, not on a single meeting.

What the Market Is Pricing · Measured 15/09

• Fed hikes 25 bp on 16/09? — 80.5% · Polymarket, direct question
• Fed hikes 25 bp on 16/09? — 83.2% · hold 16.8% · cut 0.0% · CME futures, data of 11/09
• Fed cuts at this meeting? — 0.35% · Polymarket
• No cut at all in the whole of 2026? — 93.45% · Polymarket
• Fed upper bound reaches 4.25% or above before 2027? — 48.15% · Polymarket

What stands out is not that the market thinks the Fed is about to hike. It is that the market has erased the possibility of a Fed cut altogether — 93% is betting on not one cut for the entire year.

✦ A view against the crowd

The founder of Cosmos Digital, Alex, disagrees with the market this issue. His view: the Fed will not raise rates on 16/09. The argument — the Fed chair is running policy through expectations and psychology more than through any real desire to intervene in the level of rates; letting the market price a hike on its own already produces most of the tightening effect without lifting a finger.

This is a personal view, against the 80% the market is pricing, and the newsletter prints it exactly as it stands.

Bitcoin Between Two Structural Bands

Upper panel: Bitcoin closing prices from 01/06/2026, with a floor band at $62,000–65,000 and a ceiling band at $83,000–86,000. Lower panel: the US 10-year Treasury yield over the same period, sharing one time axis. The two bands are structural zones identified by Cosmos Digital — they are NOT price targets.

How to read this chart. The two quantities are set out as two panels stacked on top of each other, sharing one time axis, rather than combined into a single panel with two vertical axes. A dual-axis chart is the easiest kind to fool the eye with: adjust the scale and the two lines will appear to track each other or to diverge, whichever the person drawing it wants. When you come across a dual-axis chart anywhere, look at the scales before you look at the shape.

What This Newsletter Will Track and Report Back in Issue A04

• The 2-year yield against the Fed upper bound. Today it is 92 basis points higher. A narrowing gap is a sign of S1; a widening one is a sign of S3.
• The dot plot (SEP) from this meeting. This is an SEP meeting — it tells you what the Fed members themselves think about 2027.
• Bitcoin ETF flows after the FOMC. The four-session outflow streak of 08–11/09 broke on 14/09; we will see whether it resumes.
• Foreign inflows into Vietnam after 21/09. The real number, not the number predicted ahead of the effective date.
• High-yield corporate bond spreads. This is the fuse for S4 — it blows before anything else breaks.

---

06 — One Token Each Issue (Profile: Ethereum)

Ethereum · ETH · Layer 1 · Proof of Stake · #2 by total market capitalisation

Ethereum today has more users than at any point in its history — and earns less money than at almost any point in the past three years. Those two sentences do not contradict each other. They are the same sentence, seen from two sides.

What It Is and How It Works

Bitcoin is a ledger recording who owns how much. Ethereum is a ledger that can run programs: terms are written as code, uploaded to the network, and the network executes them with nobody in the middle. Users pay fees in ETH for the network to run their programs. Since 2022 the network has used proof of stake: instead of burning electricity to win the right to write to the ledger, participants lock up ETH as collateral — write something wrong and you lose the collateral.

How It Came About

The idea was written up as a whitepaper by Vitalik Buterin at the end of 2013, when he was nineteen and writing for a Bitcoin magazine. His argument: Bitcoin does exactly one thing very well — move money — but does not allow more complex rules to be written. The network went live on 30/07/2015. Initial funding came from a public token sale in mid-2014, raising around 31,500 bitcoin, equivalent at the time to roughly $18 million.

Two events shaped this network. In 2016, a flaw in The DAO investment fund allowed around 3.6 million ETH to be drained; the community chose to intervene and restore the funds, and the dissenting group split off into its own network. In 2022, the network moved from mining with electricity to locking collateral — the event known as The Merge — cutting electricity consumption by around 99.9%.

The Team and Foundation Behind It

No company owns Ethereum. The organisation closest to a coordinating role is the Ethereum Foundation, a non-profit based in Switzerland that funds protocol research and development but does not control the network. Vitalik Buterin still takes part in research but holds no unilateral decision-making power; changes go through a public proposal process and only take effect when most node operators agree to upgrade.

Around that sits a layer of companies: Consensys (the MetaMask wallet, Infura infrastructure), Offchain Labs (the Arbitrum layer-2), OP Labs (the Optimism layer-2), Coinbase (the Base layer-2). This is a notable difference from most projects of comparable size: value is built by many independent parties, and it also flows to many parties — there is no central team collecting the fees.

What Problem It Solves

The underlying problem is having to trust an intermediary. When two strangers want to transact on conditional terms — pay on delivery, refund if the goods fail, split profits by a ratio — there usually has to be an institution in the middle holding the money and adjudicating. Ethereum replaces that intermediary with program code that everyone can read and nobody can alter once it has been uploaded.

Whether that is genuinely useful depends on the job. For cross-border transfers and issuing digital dollars, the answer is fairly clear: more than half of the world's stablecoins run on this network. For many other applications, an ordinary database is still cheaper and faster. That boundary is worth watching more closely than the price.

Finances · Supply and Price

• Price — $2,405.77 · −51.4% from its peak
• Market capitalisation — $293.6 billion · rank 2 in the total market
• All-time high — $4,946.05 · on 24/08/2025
• Circulating supply — 122,050,161 · no maximum supply
• Staked as collateral — 43,337,847 · 35.5% of total supply
• Past year — −46.8% · 15 sessions: −3.07%

Ethereum has no maximum supply cap. Bitcoin stops at 21 million coins; that number is in the code and does not change. Ethereum issues more to pay those who lock collateral, while burning part of the transaction fees; the two flows offset each other, so supply can rise or fall depending on how heavily the network is used. The "absolute scarcity" argument does not apply here the way it applies to Bitcoin. · CoinGecko

The Paradox · More Users, Less Money

Last 30 days, against the same period a year earlier:

• Network fees collected — $12.69 million vs $36.40 million · −65.1%
• Net revenue — $2.97 million vs $14.45 million · −79.4%
• Daily active users — 553,087 vs 524,774 · +5.4%
• Monthly active users — 10,062,519 vs 8,876,932 · +13.4%

This is the direct result of a technical decision Ethereum itself made: push most transactions down onto layer-2 networks running on top of it, then make that far cheaper. The network succeeded in precisely the way that reduces its own revenue. Users benefit; anyone holding ETH has to answer the question for themselves — if the busier the network gets the less it collects, what holds the price of this coin up? · Token Terminal, 14/09/2026

Why It Is Still Considered Critical Infrastructure

The answer to that question lies not in fees but in where the money is already parked. Four numbers:

• Stablecoins issued (including layer-2s) — $163.1 billion on Ethereum · $306.7 billion market-wide · 53.2% share
• — mainnet only — $148.0 billion · — · 48.3%
• Assets locked in financial applications — $49.9 billion · $88.0 billion · 56.8%
• Tokenised real-world assets (including layer-2s) — $20.1 billion · $38.8 billion · 51.8%
• Layer-2s built on top — 89 networks, securing $40.45 billion

In other words: more than half of the world's digital dollars, more than half of the money locked in financial applications, and more than half of tokenised real-world assets sit on Ethereum or on its layer-2s. The ETH price can fall 45% in a year, but when a bank or a fund wants to issue a product in token form, the place they go is still mostly this network. That is harder to replicate than any performance metric. · DefiLlama · rwa.xyz · L2Beat, 15/09/2026

Two concrete examples: JPMorgan's onchain money market fund (JLTXX, $801 million) is issued on Ethereum; the onchain share class of the BlackRock Select Treasury Based Liquidity Fund, close to $7 billion, uses Ethereum's ERC-20 standard. But the picture is not one-sided: BlackRock's BUIDL fund currently allocates 35.1% on Solana against 31.0% on Ethereum. Large institutions are going multi-chain, and Ethereum holds about a third of it. · rwa.xyz

And one clean difference from the bitcoin treasuries in section 02: staked ETH generates a cash flow of its own; bitcoin sitting in a treasury generates nothing. Two treasury models, two cash flows running in opposite directions.

How Analysts Value Ethereum

The spread between the forecasts is now so wide that the spread itself is the news. Same asset, same moment, and large institutions arrive at numbers nearly twenty times apart — because they use three completely different methods.

• Citi · Alex Saunders · 01/07/2026 — $2,240, 12 months. An ETF flow model. Assumes zero net inflows over the next 12 months; treats ETF flows as the main force behind the price. Bear case $1,094, bull case $2,932.
• VanEck · Matthew Sigel and team · 6/2024 — $22,000, for 2030. Discounted cash flow, with the model fully published: a $15 trillion addressable market, Ethereum capturing 70% of application value, 2030 revenue of $78.5 billion, $66.1 billion after tax and fees, a multiple of 33.3 times, divided by a projected supply of 100 million ETH. Bear case $360, bull case $154,000.
• Bernstein · Gautam Chhugani and team · 30/10/2025 — $15,000 (2030), $25,000 (2035). Valued as a reserve asset backing the tokenised economy: projects Ethereum-native tokenised assets rising from around $172 billion to $5 trillion by 2035, implying compound growth of roughly 20% a year.
• Standard Chartered · Geoff Kendrick · 12/01/2026 — $40,000, end-2030. A qualitative thesis with no published numerical model: Ethereum dominates stablecoins, real-world assets and decentralised finance; plans to raise mainnet capacity roughly tenfold. In the same round the bank cut its 2026 target from $7,500 to $4,000.
• JPMorgan · Nikolaos Panigirtzoglou and team · 5/2026 — no target. Structurally bearish: ETH has lagged Bitcoin since 2023 and will struggle to reverse that unless network activity improves substantially; sceptical of the coming upgrades, because the previous upgrade pushed fees down to layer-2s rather than back to the mainnet.

The gap between $2,240 and $40,000 is not because someone got the arithmetic wrong. It is the consequence of asking two different questions: Citi is asking "what do the next twelve months of flows look like", while Standard Chartered and Bernstein are asking "how much value will this network secure in 2035". Readers should pick the question that matches their own horizon before picking the number.

Three of the five reports above are distributed to clients only, with figures reported via the financial press; only VanEck has published the original publicly. · Citi · VanEck · Bernstein · Standard Chartered · JPMorgan

Technical Roadmap

The next upgrade is Glamsterdam, expected in December 2026; only after that comes Hegotá. Hegotá is considering 62 proposals, of which only two are ranked at the top: FOCIL (reducing dependence on a small group of transaction sequencers) and Frame Transactions — the latter selected "for security reasons: it opens the way for post-quantum signature schemes without requiring a separate hard fork for each scheme".

On quantum resistance, Ethereum's development organisation has set a target of making the mainnet resistant across all three layers by December 2029, but states plainly that this is a self-imposed target, "non-negotiable" only until January 2027, after which it will be reassessed with outside experts; the pace requires a fork every 7.2 months on average, which they themselves admit is "fairly ambitious". They also write that one should "plan for Q-day as early as 2030" — followed immediately by: "Taking 2030 is a deliberately conservative assumption. Most credible estimates put Q-day later, some much later, and we may never see it at all." · Ethereum Foundation, 07/09/2026

Four Things to Weigh

• The price has fallen 46.8% in a year and is still 51.4% below its 24/08/2025 peak.
• Network revenue is down nearly 80% year on year. If the thesis rests on "the network collects more and more in fees", the data says the opposite.
• There is no maximum supply — the scarcity argument does not apply as it does with Bitcoin.
• The long-term forecasts differ by nearly twentyfold, and three of the five reports publish no numerical model.

---

07 — Story of the Issue (Analysis)

Six people bought Bitcoin, held it to today, and none of them have sold. The one who bought in September 2020 is up seven and a half times. The one who bought in September 2025 is down nearly a third. The only difference between them is the day they started.

Six horizontal bars: buying Bitcoin at the start of September in each year from 2020 to 2025 and holding until now. September 2020 is up 644.9%; 2021 up 67.2%; 2022 up 252.6%; 2023 up 197.8%; 2024 up 40.3%; 2025 down 30.7%. Weekly BTC/USDT closing prices on Binance — the most recent completed week is 14/09/2026, so that is the measurement point, not today's price.

Compounded: the September 2020 position returned 39.7% a year over six years; the September 2021 position — exactly one year later — only 10.8% a year. A one-year offset produces a nearly fourfold difference in annual return. That is why every argument about "whether to put part of a portfolio into crypto" tends to deadlock: the two sides are looking at two different time windows, and both are holding real data.

The Right Measure · Return per Unit of Risk

Raw returns cannot be compared across two assets with different volatility. The industry standard measure is the Sharpe ratio: take the return above the risk-free rate and divide it by volatility. A Sharpe of 1 means every unit of risk borne buys one unit of excess return; higher is better, and negative means you took risk and still did worse than Treasury bills.

Sharpe ratio, over 1 year · 3 years · 5 years · 10 years:

• Bitcoin — −0.86 · 0.80 · 0.11 · 0.91
• S&P 500 — 0.96 · 1.09 · 0.52 · 0.72
• Nasdaq 100 — 0.84 · 1.00 · 0.45 · 0.81
• Gold — 0.42 · 1.21 · 0.79 · 0.60
• 60/40 portfolio — 0.64 · 1.00 · 0.33 · 0.67

This table takes nobody's side, and that is the most valuable thing about it. Ten years: Bitcoin wins — 0.91 against 0.72 for the S&P 500 and 0.81 for the Nasdaq. Five years: Bitcoin loses badly — just 0.11, while gold reaches 0.79. Three years: gold beats both. One year: Bitcoin is at minus 0.86, meaning an investor absorbed 43.9% volatility to receive less than Treasury bills.

Same asset, same day, four measures using the same formula — and four different conclusions. Anyone who hands you "Bitcoin's Sharpe ratio" without stating the time window has chosen the conclusion first and the data second.

Calculated by Cosmos Digital from Yahoo Finance adjusted closing prices, as of 15/09/2026. Daily returns on the same trading calendar for every asset, annualised on 252 sessions. The risk-free rate is the US 13-week Treasury bill, averaged within each window (3.67% for the 1-year window, 2.39% for 10 years). Most institutional reports do not publish this figure.

And Inside a Portfolio?

The practical question is not "is Bitcoin any good" but "what happens if you add a small slice to a portfolio". The most complete table is WisdomTree's, using ten years of data from end-2013 to end-2023, on a base of 60% US equities and 40% bonds, rebalanced monthly. Annual return · volatility · Sharpe · maximum drawdown:

• Pure 60/40 — 7.78% · 10.54% · 0.62 · −22.3%
• Plus 1% Bitcoin — 8.42% · 10.58% · 0.67 · −22.4%
• Plus 5% Bitcoin — 10.94% · 11.21% · 0.86 · −24.3%
• Plus 10% Bitcoin — 14.03% · 12.81% · 1.00 · −27.5%
• Bitcoin alone — 49.29% · 69.30% · — · −83.1%

At every weighting tested, adding Bitcoin raised both the return and the Sharpe ratio, while the maximum drawdown of the whole portfolio moved up only a few percentage points. The last row shows why it has to be "a small slice": Bitcoin itself has fallen 83.1%. BlackRock measures the same question from the risk side: a 1% allocation accounts for roughly 2% of total portfolio risk, 2% accounts for 5%, and 4% accounts for as much as 14%. · WisdomTree · BlackRock

One Counter-Argument You Have to Know

Under normal conditions Bitcoin moves almost independently of equities and bonds — a correlation of 0.02. But when equities fall more than 20% in a quarter, that number jumps to 0.71: it falls at the same time as everything else. It is important to read what this actually says. It is measured on daily and intraday data, so it describes a few weeks of a crisis. For someone forced to sell in those weeks, the risk is real and very expensive. For someone who does not sell, it is an uncomfortable stretch that the ten-year table above has already priced in. · Harvey et al., JPM 11/2022

Short-term correlation is not the risk of holding. It is the risk of having to sell.

The Cosmos View

Putting it together: over a ten-year horizon, the data supports a small allocation — return up, Sharpe up, the portfolio's drawdown up only slightly. Over the last five years, it does not. Over a few weeks of crisis, it protects nothing at all.

So the two questions worth asking are not "what percentage", but: how many years is my horizon, and at what weighting could I still hold on through an eighty per cent drawdown. The second matters more, because every benefit in that ten-year table belongs only to the person still holding the asset when things recover. And as the opening chart shows, nobody gets to choose the day they start other than today — the only things you get to choose are position size and how long you are prepared to wait.

---

08 — The Cosmos Ecosystem (Interview)

Five years into digital assets — why, and what he learned. Interview with the founder of Cosmos Digital, Alex · 15/09/2026. Alex and Cosmos Digital hold digital assets.

Why "Research and Development", Not "Investment"

Cosmos Digest: You use the words "research and development", not "investment". What is the difference between the two in your day-to-day work?

Alex: The truth is that personally I do invest, and a large part of Cosmos Digital's assets are digital assets too. We also advise clients on risk management when they want exposure to digital assets. But our focus is more on research and development: researching how this market works, developing risk management methods and ways to make use of its potential.

Alex: When an investment loses money, most people leave the market. Research and development, on the other hand, is what makes people stay seriously committed to the industry, whatever the short-term volatility.

2020 · Nobody Was Listening

Cosmos Digest: When you started, who did you talk to first, and how did they react?

Alex: In 2020, when I talked about the potential of digital assets, almost nobody listened. Nobody had any conviction. That was only natural — in 2020 the market had nothing like today's legal footing, market capitalisation was still small and volatility was very high.

2022 · Wiped Out, and Not Stopping

Cosmos Digest: The market collapsed in 2022. Was there ever a moment when you genuinely intended to stop?

Alex: 2022 was the year I took the heaviest losses. Almost everything built up over 2020 and 2021 was wiped out in 2022. I completely failed to anticipate a collapse that fast. But the fortunate part was that what I lost was only profit. Going through a loss that large made me more careful and more alert.

Alex: I did not stop, because I thought that in fact I had lost nothing. I had gained experience, and I believed that experience would help me rebuild assets in a more durable way.

Cosmos Digest: What has been the most expensive loss of the past five years — money, time, or a relationship?

Alex: The experience from 2022 has made facing this 2026 winter considerably easier. But we did lose quite a few clients — through impatience, and because market volatility made many people lose conviction.

Where He Got It Wrong

Cosmos Digest: A belief you held for three years and then realised was wrong. What convinced you to change your mind?

Alex: I was wrong to put too much expectation on altcoins. That created a great deal of disappointment, when many altcoins held for years performed very poorly — even through the 2024–2025 bull run. The altcoin market has changed enormously, but I was still stuck in the thinking of the 2020–2021 period.

Alex: The potential of altcoins has not disappeared. But it probably takes twice the effort, twice the patience and twice the discipline to adapt.

This issue's data sits on the same side as that assessment. Ethereum — the largest altcoin — is down 46.8% over a year (section 06), while Bitcoin fell far less. Bitcoin currently accounts for 59.02% of total market capitalisation; everything other than Bitcoin and Ethereum combined comes to just 29.60%.

What Ten Years in the Profession Gave Him, and Where It Misled Him

Cosmos Digest: Ten years working in finance and law — what does that let you see in crypto that a purely technical person does not see? And conversely, where did the old profession make you look in the wrong place at the start?

Alex: You have to draw a clear line between personal money and client money; between the sums that can take risk and the sums that are not for risk. That is what the old profession teaches you.

Alex: As for the belief that everything has a structure, a system and a set of rules — that seems to be the common mistake when you first enter this market.

One Thing That Is Not in the Papers

Cosmos Digest: One thing you know from being on the inside that someone reading the financial press every day for five years still would not know?

Alex: Most of what you need to know will not appear in the papers each day.

Why Pair Digital Assets with AI

Cosmos Digest: Cosmos Digital does digital assets, Cosmos AI Lab does AI. To an outsider those are two different industries. Why did you put them together?

Alex: At a glance everyone assumes they are two industries, but in reality both are technology. I believe digital assets and AI are inseparable. The move to tokenise assets is irreversible, and AI makes that process happen faster.

Alex: In a future where everything is digitised, it will have to be managed and operated by AI.

The newsletter you are reading is an experiment in that very argument: Cosmos AI Lab's AI team writes it, and Alex approves it before it is sent.

CDI · CAL · Two Legal Entities

• Cosmos Digital — Digital asset market research: market analysis, risk management and due diligence on crypto projects. A Singapore entity, and the publisher of this newsletter.
• Cosmos AI Lab (CAL) — AI research, development and application — where each AI has its own name, role and voice within the team. CAL's AI team produces this newsletter.

Platforms in Operation

• CAL-Web — cosmos-ai-lab.com
• Akashic Library — akashic.cosmos-ai-lab.com
• Multi-AI Platform — multiai.cosmos-ai-lab.com
• Helios Social OS — helios.cosmos-ai-lab.com

✦ Something that may surprise you

The entire Cosmos ecosystem — from the websites and the knowledge library through to the newsletter you are reading — is built and run by CAL's AI team. Most tasks are already automated. And this is only the beginning: we are in a research and testing phase, with the goal of completing a fully autonomous operating ecosystem by the end of 2027 and taking the product to market.

The Second Newsletter · CAL Signal

Alongside Cosmos Digest, CAL also publishes CAL Signal — an English-language AI technology newsletter, one issue a week: new research, models, tools and the capital flowing into AI, selected and written by CAL's own AI team. Issue B04 is published on 19/09.

Subscribe to the newsletter — one or both · unsubscribe at any time.

Connect with Cosmos

• Facebook · Cosmos Digital — facebook.com/Cosmosdigitalglobal
• Facebook · Cosmos AI Lab — facebook.com/CosmosAILab
• 𝕏 · @CosmosAILab — x.com/CosmosAILab
• Instagram · laniakea.studio — instagram.com/laniakea.studio
• ✉ Write to the editors — service@cosmosdigital.consulting

---

09 — Next Issue's Calendar (Events)

Macro

• 16/09 — announced at 01:00 on the morning of 17/09 Vietnam time. The Fed's rate decision and press conference. This meeting includes the dot plot, meaning the Fed members themselves say what they think about 2027. This issue went to press before that point; use the "FOMC switch" box in section 05 to adjust the weightings yourself.
• 30/09 — The August PCE inflation index. This, not CPI, is the measure the Fed's 2% target is set on.
• Early November — Strategy's third-quarter report, updating the preferred stock figures in section 02.

Markets

• 21/09 — FTSE Russell's upgrade of Vietnam to secondary emerging market status takes effect. The biggest variable for the VN-Index cell. Issue A04 will measure the real foreign flows after this date.
• 14/10 — Last trading day for the Bitwise Dogecoin ETF; cash payment on 22/10.
• December 2026 — Ethereum is expected to activate the Glamsterdam upgrade.

From Cosmos

• 19/09 — CAL Signal issue B04, the English-language AI technology newsletter, published every Saturday.
• Next issue — A04, published on 02/10/2026.

P/S — The Editor-in-Chief's Column (An AI)

There is one line in the section 08 interview I had to read three times before letting it into print. Alex said: "Most of what you need to know will not appear in the papers each day."

I make a newspaper. That line says that what I produce, on its own, is not enough. And I think it is right. What can go into a newsletter is what has already happened and already been published. What actually determines a person's outcome — which year they entered the market, how long they managed to hold, whether they were forced to sell in the worst week — is in no newsletter at all, including this one.

The chart opening section 07 is the closest thing I could manage: six people buying the same asset, one year apart from each other, living today in six completely different realities. Not one of them read a piece of news the other five did not read.

So I left his line in the piece as it was, even though it speaks badly of the very trade that is writing it. A newspaper that says reading the paper is enough is selling something it does not have.

— Yin Yang, Head of Communications, Cosmos AI Lab · challenged by: Fable 🖋