Three Names at Five Billion Each: Inside Asia-Pacific's AI Fundraising Cycle

In the final quarter of last year, any equity desk in Raffles Place or Central would have shown you the same scene. Three names on a whiteboard — Firmus, DayOne, Yangtze Memory. Beside each, the same-sized number: roughly five billion dollars apiece. The air in the room dry, voices low. There was nothing to shout about. Putting three deals of that size on the pricing table at once is confidence and arithmetic at the same time.

That same week, a sentence surfaced in the press that did not quite fit the whiteboard. The arranging bank itself said investor selectivity was rising. The market being headlined as racing toward a record was, from the inside, quietly narrowing its door.

I have spent years writing down the arithmetic of sport — who ran how far, who carried what load. The arithmetic here is the same shape. One difference: here the companies run, and the investors carry.

The numbers, and the number nobody counts

Asia-Pacific equity capital markets raised $327.1bn in the first nine months of the covered year, up 53% year on year on LSEG and Dealogic data. The source text is inconsistent about the year — sometimes "this year", sometimes "the first nine months of 2026". Every figure below follows the source and requires verification.

High technology supplied $125.8bn of that, 38% of the total and more than three times the prior year. It sounds simple. The arithmetic is not. Roughly one and a half of every four dollars is flowing into a single story — AI-linked chips, data centres and power.

The 2026 comparison matters. That year's full-year Asia-Pacific total was $557.6bn. To break it, the current year needs $230.6bn in the fourth quarter alone — a single-quarter figure never achieved before.

The headline says "on track for a record". The internal arithmetic says "conditionally on track". The gap between those two sentences is what this piece is about.

Where the money comes from, where it goes

In 2026 the training ground got Wi-Fi, and the silence between drills got shorter. That year I logged 47 sessions and the sleep, load and recovery of 22 players. The lesson was simple: when a system speeds up, it shows first in the length of the pauses, not in the results.

In markets, that pause is the roadshow. The fewer empty hours between meetings, the faster the machine. Last year the gap approached zero.

The transmission chain is straightforward. At the top, demand for AI compute. In the middle, equity financing. At the bottom, that money becomes chips, data centres and power lines. LSEG and Dealogic show the middle layer at its busiest in history. Goldman Sachs' view is that AI will keep driving market volumes for the next one to two years.

The middle layer is the most volatile, because that is where money arrives on a promise — and promises move first. A chip order book takes six months to change. A power plant takes three years. A deal can be pulled in one morning.

The venue war: Hong Kong, Mumbai, Nasdaq — and Singapore's narrow door

More striking than the totals is who is winning. Hong Kong is recovering large listings. Mumbai is rising up the league table of active issuance markets. And Asian companies that feel undervalued at home are going to Nasdaq.

SK Hynix's $26.5bn share sale on Nasdaq is the marquee thermometer. A Korean chip company raising that scale in the United States signals demand — and something else: the region's own investor base cannot absorb an issue of that size alone.

Singapore's position is different, and to my eye the most interesting. The city cannot be a major listing venue itself, yet it is supplying names to the pipeline — DayOne, a deal of roughly $5bn. The city losing the listings is now manufacturing the product that gets listed. Venue arithmetic and company arithmetic have come apart.

There is a sporting parallel. A league cannot keep its best players, but it can keep producing them. Venues now sit in exactly that position: the product stays, the stage does not.

The names in the pipeline

Names say more than totals. Firmus in Australia, DayOne in Singapore, Yangtze Memory in China — all around the $5bn mark, all technology infrastructure. Alongside them, Mynt in the Philippines, Samsung Biologics in Korea, Reliance Jio in India. The list is not monochrome, but technology sits at the centre.

High technology is already 38% of issuance. The pipeline suggests that share rises further.

One thing goes unsaid on the pricing call. An issue pricing means money raised. It does not mean infrastructure built. Data-centre approvals, land, grid connections, cable — none of these run on the timetable of a share sale.

A transfer fee is just a number until you watch a nineteen-year-old pack his bags. Issuance volume is the same: just a number until you see the worker pulling cable into a data centre, or the electrician standing at the substation gate. That labour never appears in a prospectus, because nobody is issued shares for it.

The metric everyone watches, and the one nobody does

In sport there is a habit I have seen many times. Kilometres covered, sprints made — printed after every match. The numbers look good. The problem is that pointless running also produces pretty numbers. A player chasing back is still accumulating kilometres.

Issuance volume is that kind of metric. $327.1bn looks magnificent. It does not say which money went to good use, who bought at what price, or who will be in profit four years from now.

The structure is shifting as the number grows. Not just IPOs — follow-ons, rights issues, convertible bonds. And convertible bonds and private placements do not attract the same daylight scrutiny a full public offering does.

Money that never enters the prospectus template never enters the market's story — but the market still carries its weight. Judging any single deal means looking not only at the size but at how open the door was.

The record rests on a condition

The 2026 record was $557.6bn. Nine months in, the current year holds $327.1bn. Closing the gap requires $230.6bn in the final quarter. No quarter has ever produced that.

Not impossible. If the pipeline names all reach the pricing table, the arithmetic works. But "if" is the operative word, and headlines do not carry it.

The second point matters more. The 2026 record and this one are not the same kind of record. That year was a blend of technology and SPACs. This year is one theme: AI. When the composition changes, the comparison must change too.

The Japanese fans were still singing in the 94th minute, one goal down. What I learned that evening is that the ones who stay in the final minute are the real arithmetic. In issuance, that final minute is the fourth quarter. The three $5bn deals not yet priced will tell us how open the door really is.

The caution from Citigroup — rising investor selectivity — is this cycle's first brake. It is still soft. Brakes always arrive soft, then suddenly firm.

There is another reason this matters. The first to leave a pipeline are the marginal issuers, the ones with thin stories and stretched valuations. The big names get funded in the first pass. The damage lands on the next layer, where smaller companies and ordinary investors stand.

What to watch in the final quarter

First, whether Firmus, DayOne and Yangtze Memory price — and at what level. Pricing proves depth. Delay or withdrawal proves supply has outrun demand.

Second, the high-tech share. If 38% climbs further, both the upside and the risk climb with it. A market standing on one theme looks elegant until it stumbles.

Third, the language of bankers. When "some caution" becomes "selectivity has risen", the cycle is approaching its peak. Market tops never announce themselves; they arrive as one sentence replacing another.

And finally, the capital plans of chip, data-centre and power companies. This entire fundraising cycle rests on a single assumption — that AI-linked spending will not slow over the coming years. That assumption will change one day. The question is not whether, but how much money is already committed before it does.

Glossary

ECM (Equity Capital Markets): the investment-banking function handling share issuance, IPOs, follow-ons and convertible bonds.

Follow-on: a secondary share sale by a listed company.

Convertible bond: a bond that can later convert into shares.

IPO: a company's first public sale of shares.

Rights issue: new shares offered to existing shareholders, usually at a discount.

LSEG: London Stock Exchange Group, a major market-data provider.

Dealogic: a capital-markets data and analytics provider.

AI capex cycle: the wave of corporate capital expenditure on AI-linked chips, data centres and power.