SEA's Late-Stage Surge: Five Deals That Tell the Reallocation Story

Originally published: 2024-05 | Last verified: 2026-09-13 Round sizes and lead investors are taken from each company’s own announcement, linked inline; where a valuation comes from press coverage instead, the text says so. An earlier version of this piece misstated the Airwallex and EPG round sizes, described EPG as a fintech, mischaracterised Amity’s business and then its geography, described three rounds as flat that were not, named participants as leads, and said strategic investors came in as participants rather than leads, which the announcements do not show; earlier versions also left out Airwallex’s December 2025 Series G, which falls inside the window. Those are corrected here, and the totals, shares and lead counts below include the Series G. The window covered is April 2025 to March 2026. Funding round details revise as further disclosures land; please confirm against primary source coverage for current deal terms.

The “SEA late-stage surge” line on Tracxn-derived charts since H1 2025 is often asked to carry more of the story than one line can. Charts compress; they show late-stage funding rising while seed-stage funding falls. The story — what kind of capital came back, what flavour of company it backed, what the underlying thesis was — gets lost in the compression.

In my pillar piece on SEA reallocation I argued that the right framing for the post-2022 environment is reallocation rather than recovery. This piece looks at that thesis at the level of specific deals: five companies’ rounds between April 2025 and March 2026. They are not representative of every deal, and they do not describe the whole late-stage book; they are a sample, chosen because they illustrate recognisable patterns, and what follows is a reading of that sample.

I’ve picked these five because they cluster around recognisable patterns rather than because they were the largest. A caveat I got wrong first time round and want to state plainly: I originally set the DayOne USD 2B round aside as a special case — data centre infrastructure, financed against contracted utilisation revenue, closer to project finance than growth equity — and then included EPG in the “conventional growth-stage book.” EPG is also data centre infrastructure. So here is the honest version. The sample counts every round each company raised inside the window: five profiled companies (seven rounds, since Airwallex and EPG each raised twice) plus DayOne for scale — six companies, eight rounds, about USD 3.2B in all. By dollar value, the two data-centre businesses, DayOne and EPG, are about two-thirds of it (USD 2.2B). By count they are two companies out of six and three rounds out of eight. In this sample, data centres are not the exception to the late-stage story in dollar terms; in deal terms they are one pattern among several.

The five deals, one paragraph each

Thunes — USD 150M Series D, April 2025. Cross-border payments infrastructure. Led by Apis Partners and Vitruvian Partners, and closed at what Thunes describes as “a substantial valuation increase over its last round” — the 2023 Series C, which had been led by Marshall Wace with Bessemer participating, at a valuation of over USD 900M. DealStreetAsia put the Series D valuation at USD 1.42B.

Airwallex — USD 630M across two rounds, May and December 2025. Multi-currency global business banking platform, founded in Melbourne and headquartered in Singapore; DealStreetAsia counts its rounds in its Southeast Asia tallies. The May Series F totalled USD 300M at a USD 6.2B valuation, of which USD 150M was secondary share transfers — worth separating, since only the primary half is new capital into the business. No lead was named; participants included Square Peg, DST Global, Lone Pine Capital, Blackbird, Airtree, Salesforce Ventures and Visa Ventures. In December came a USD 330M Series G at a USD 8B valuation, led by Addition with participation from T. Rowe Price, Activant, Lingotto, Robinhood Ventures and TIAA Ventures; Addition is a New York-based venture capital firm, and the same announcement made San Francisco a second global headquarters alongside Singapore.

bolttech — USD 147M Series C, June 2025. Insurtech infrastructure and embedded insurance distribution. Shares acquired by Series C investors totalled USD 147M at a USD 2.1B company valuation. No lead was named. The announcement welcomes Sumitomo Corporation, the Japanese integrated trading and business investment company, and Iberis Capital, which the release describes as a leading private equity and venture capital manager in Portugal, as strategic investors, and says they join Dragon Funds, Baillie Gifford, Generali’s Lion River and other investors in the round. Against the USD 1.6B Series B of May 2023, that is a 31% up-round, not the flat mark I previously carried.

EPG — USD 200M across two Series B rounds, Q1 2026. Not a fintech, as I had it: EPG is a Singapore-headquartered modular data centre manufacturer, doing factory-prefabricated integration of power, IT and cooling systems, with R&D centres in Singapore and Shanghai and manufacturing in Malaysia and China. The January round, of nearly USD 100M, was co-led by Forebright and Silicon Peak; the March Series B+, of over USD 100M, was, in the release’s words, “led by Decarbonization Partners, a joint venture between BlackRock and Temasek, Alibaba Cloud, and other strategic investors”. The release puts all three in one “led by” list and does not separate leads from participants. Tracxn counts the two together as USD 200M. One definitional note: Tracxn’s stage definitions put Series A and B rounds in early-stage and Series C onward, PE and pre-IPO rounds in late-stage, so on Tracxn’s count these two rounds are early-stage. I keep them in the sample because I am using “late-stage” for large rounds into companies well past the start-up phase — EPG, per its January release, was founded in 2004 — rather than by series letter; every other round in the sample is Series C or later. Drop EPG and the sample is about USD 3.0B, with DayOne alone about two-thirds of it; the dollar picture barely moves, but Pattern 4 would lose two of its six rounds with a named lead, one of them state-linked.

Amity Solutions — USD 100M Series D, March 2026. A Thailand-founded software and AI technology group, raising to accelerate vertical and agentic AI. The round was led by EDBI, the investment arm of SG Growth Capital, alongside Asia Partners and SMDV, with CMLIM Capital among the participants. Amity describes it as the largest GenAI-focused round in Southeast Asia to date. Thailand-founded does not mean Thailand-bound: the same announcement says more than 75% of Amity’s 2025 EBITDA came from its European business units, that its AI research centre, ARAC, is headquartered in Singapore, and that it is expanding across Southeast Asia and Europe. So this is a Thailand-founded group earning most of its profit in Europe, with a Singapore state-linked investor leading.

These five companies’ seven rounds together come to roughly USD 1.2B between April 2025 and March 2026; add DayOne and the sample is about USD 3.2B. They do not represent the full late-stage book — DealStreetAsia counted 10 late-stage deals in H1 2025 and 24 in H2, 34 for the year on my sum — but I picked them because they cluster on recognisable patterns, so read the patterns below as a reading of this sample rather than of the whole late-stage market.

Pattern 1: The capital came back to infrastructure, not to consumer

The first read across the five deals is that the late-stage capital in this sample went to infrastructure-shaped businesses, not to consumer-shaped ones. Thunes and Airwallex are payments and currency rails. Bolttech is embedded insurance distribution. EPG is data centre infrastructure. Amity is enterprise application infrastructure. None of these are consumer marketplace plays, none are e-commerce extensions, none are food-delivery or ride-hailing variants. My impression of the 2017-2021 SEA late-stage book is that its best-known rounds went to consumer platforms (Grab, Sea, Tokopedia, Gojek), but I have no sourced breakdown of that period’s late-stage dollars by sector — the same gap that stops me drawing a lead-investor comparison in Pattern 4 — so treat the contrast as my reading, not a measured shift. If this sample is indicative, the difference points to what kind of unit economics late-stage capital is now willing to underwrite.

On my read, infrastructure businesses at this stage tend to have more predictable revenue per customer, more multi-year contracted revenue visibility, and lower customer acquisition costs as a percentage of LTV than the consumer marketplace generation did. They tend to underwrite with lower steady-state operating margins but clearer gross margins and much lower CAC volatility. For late-stage capital that learned the wrong lessons from the 2021 consumer marketplace exuberance, that profile would be more comfortable to underwrite — growth and crossover funds can model the next twelve quarters with a tighter error bar than they could on a consumer-marketplace target. That description fits the payments, insurance and software businesses more cleanly than EPG, which is a modular data centre manufacturer; on my read its revenue follows units built and delivered for data-centre projects rather than a recurring per-customer fee, so it belongs in this pattern for what it sells into rather than for how it earns.

The implication for LPs, if the sample is indicative: the late-stage SEA book being built now may not have the same business mix as the consumer-heavy book LPs remember from 2018-2022 — a contrast I can only draw impressionistically, as above. Underwrite it deal by deal rather than by analogy with the last cycle.

Pattern 2: The capital came back to companies with global revenue mix, not to SEA-only ones

The second read is geographic. Thunes, headquartered in Singapore, says its network spans more than 130 countries, with offices from Nairobi and Johannesburg to London and San Francisco. Airwallex, now co-headquartered in Singapore and San Francisco, holds licences across North America, Europe, the Middle East and Asia-Pacific. bolttech, as of its 2023 Series B, served customers in more than 30 markets across North America, Asia and Europe. EPG, according to its March release, serves “hyperscale, cloud, and enterprise deployments across APAC, EMEA, and other global markets,” from manufacturing hubs in Malaysia and China. Amity is Thailand-founded, but more than 75% of its 2025 EBITDA came from European business units and its AI research centre is headquartered in Singapore.

On the companies’ own descriptions, none of the five presents itself as a pure-play SEA business — one with most of its revenue and most of its customers in SEA — though only Amity discloses where its profit comes from. All five are closer to “a SEA base or SEA origin, with operations across a multi-region footprint.” In this sample, the late-stage capital went to businesses that are bookable through SEA rather than primarily of SEA.

This would matter for two reasons, both of them hypotheses rather than findings, since five companies cannot show what the whole late-stage book looks like. First, it would help explain the SG capture rate: if the late-stage book is dominated by multi-region businesses headquartered in Singapore, Tracxn-style geography reporting will keep showing high SG concentration even when the underlying operating activity is dispersed across markets. Second, if the pattern holds beyond the sample, country-level SEA fund managers (Indonesian-only or Vietnamese-only thesis funds) would see fewer late-stage candidates in their pipelines than the regional aggregate suggests, because the companies graduating to late-stage would be the multi-region ones rather than the country-pure-play ones. I have no pipeline data from such funds to test that.

We used to see one hundred SEA-pure-play candidates for every late-stage round we’d consider. Now we see twenty multi-region candidates with SG hubs. The denominator changed. The numerator that gets funded looks completely different.— SEA-focused growth fund partner

Pattern 3: The valuations that were disclosed rose

The third read is on valuation, and it is the read I most need to correct. I previously wrote that three of these rounds closed flat to the prior mark. That is not what the disclosures say. Three of the five have a prior mark on the public record:

• Thunes: its 2023 Series C valued it at over USD 900M; DealStreetAsia put the Series D at USD 1.42B — about 58% higher at most, since the earlier figure is a floor, about 21 months later.
• bolttech: USD 1.6B at its May 2023 Series B to USD 2.1B in June 2025 — 31% over two years, about 14% a year.
• Airwallex: USD 5.6B at its October 2022 Series E extension (the figure in that announcement’s headline and text; its web address says 5.5) to USD 6.2B at the May 2025 Series F and USD 8B at the December 2025 Series G — about 43% over a little more than three years, or about 12% a year. Most of it came inside the window: the last step, USD 6.2B to USD 8B, was about 29% in under seven months, which Airwallex’s December release rounds to “~30%”.

EPG and Amity did not disclose valuations; nor did DayOne, the scale reference, though its release says the Series C was priced at a 100 percent premium to the prior round. So none of the five is a confirmed flat round, and I should not have said otherwise.

What the three disclosed pairs share is direction, not a common pace: annualised, they run from about 12% (Airwallex, October 2022 to December 2025) through about 14% (bolttech) to just under 30% (Thunes, an upper bound since its 2023 mark is a floor). Three data points are not a market rate, and I have dropped the claim I used to make here about how today’s entry prices compare with 2021 — the disclosures do not support it. What they do rule out, for this sample, is the “flat round” story.

For founders in the SEA growth-stage market, the clarifying implication is narrower than I used to put it: up-rounds were available in this window, at least to businesses like the three that disclosed their valuations. For LPs, the entry-price question has to be answered deal by deal rather than assumed from the cycle. Whether the exits ultimately validate these marks is the question for 2028-2030.

Pattern 4: Who led — mostly financial investors by count, US funds by dollar

The fourth read is on who is actually in these rounds. Six of the eight rounds name a lead; Airwallex’s May round and bolttech’s do not. By count, four of the six are led by financial investors — growth-equity, venture, crossover and other investment firms: Apis Partners and Vitruvian Partners on Thunes, Forebright and Silicon Peak on EPG’s January round, Addition on Airwallex’s Series G, and Coatue on DayOne. Two are led by state-linked capital: EDBI on Amity, and Decarbonization Partners, the BlackRock-Temasek joint venture, on EPG’s March round. That March round is also the only one that puts a strategic investor on the lead side, if its release is read at face value — the only reading it offers, since it does not separate leads from participants: Alibaba Cloud and other strategic investors are in the same “led by” list as Decarbonization Partners. Elsewhere, strategic and corporate money appears in the two rounds that name no lead (Sumitomo in bolttech; Salesforce Ventures and Visa Ventures, the latter described by Airwallex as a strategic investor, in Airwallex’s May round) and as named participants in one round (Robinhood Ventures and TIAA Ventures in Airwallex’s Series G). bolttech’s release also calls Iberis Capital a strategic investor, but it is a private equity and venture capital manager, so I count it with the financial investors.

By dollar weight the picture is different. The two largest rounds in the sample were led by US-based funds: Coatue, a US crossover fund, led DayOne’s Series C, with Indonesia’s sovereign fund INA alongside, and Addition led Airwallex’s Series G. Together that is about USD 2.3B of the roughly USD 2.8B in rounds with a named lead — about 84% — and about 72% of the whole sample. One of the state-linked leads, Decarbonization Partners, is itself a joint venture with a US asset manager, BlackRock. So this sample gives no support to the idea that late-stage SEA capital is tilting away from US growth funds: by count, US-based funds led two of the six rounds with a named lead; by dollars, they led most of the money. Five companies, with no sourced picture of who led SEA late-stage rounds in 2018-2021 to compare against, cannot show a shift in either direction; I’d treat any tilt as a hypothesis, not a finding.

What the sample supports, then, is narrower than I used to claim. Where leads are named, most are financial investors, with state-linked capital leading two of the smaller rounds; by dollars, US-based funds led most of the money. Strategic and corporate money is present, but the sample cannot say whether it leans toward participating or leading: it is on the participant side of Airwallex’s Series G, on the “led by” list of EPG’s March round, and in the two rounds that name no lead at all. I have dropped two claims I used to make here — that this capital is more conservative on valuation, and more disciplined on underwriting, than the leaders of the 2021 cycle. The disclosed marks in Pattern 3 do not show the first, and nothing in these announcements shows the second.

Whether the lead-investor base behind SEA late-stage rounds is changing is the question this sample cannot answer, and it is worth watching: a different lead-investor base would bring different exit horizons, return expectations and operating involvement preferences, and the companies that fit it would differ from the ones that fit the old one.

The concentration these rounds sit inside is the subject of Singapore’s capture rate.

Pattern 5: The exits are still missing

The fifth read is what the sample cannot show: a recent comparable exit for businesses like these. Exit activity slowed in 2025, on DealStreetAsia’s count: only 57 acquisitions and 15 tech IPOs across the region, and Edgar Hardless, CEO of Singtel Innov8, quoted in coverage of the report, said the lack of exits was the single biggest drag on investor confidence. Sea Limited (NYSE) and Grab (NASDAQ) remain the reference SEA tech listings on global exchanges, and neither is a recent precedent for the kind of multi-region infrastructure business in this sample.

On my read, this is the part of the late-stage surge story that should make LPs most cautious. In this sample, capital went into multi-region, infrastructure-shaped businesses at valuations that rose where disclosed — but the realisation pathway for that capital is unproven in the current cycle. The best-known exits of the last cycle were public listings, Sea’s and Grab’s among them; the 2024-2026 vintage will need either a comparable listing route or another exit pathway (M&A by global strategics, secondary trades to sovereigns, alternative listing venues).

The pattern that worries me the most: late-stage capital deployed into multi-region SaaS infrastructure businesses with US-comparable revenue profiles, at SG-holding-company structures, where the natural exit path is a US listing and nothing in this sample shows that path open for businesses like these. That capital may need to hold for substantially longer than the underwriting horizons assume. My read, which these announcements only hint at, is that until listings for such businesses pick up, more of the liquidity in this book will come through secondary trades than through primary realisations. The one example in the sample is Airwallex, where USD 150M of the USD 300M May round was secondary share transfers: existing shares changing hands rather than new capital into the business. Its December release does not say whether any of the Series G was secondary.

What this means for reading the 2026 SEA late-stage book

Synthesising across the five-deal sample:

• Sector mix: data centres, payments, insurance, enterprise software — with data centres about two-thirds of the sample by value. Not consumer marketplaces.
• Geographic mix: multi-region across all five on their own descriptions, most of them based in Singapore; none presents itself as a pure-play SEA business, and Thailand-founded Amity earns most of its EBITDA in Europe.
• Valuation: rising where disclosed (three of five companies); none confirmed flat.
• Investor mix: financial investors lead four of the six rounds with a named lead, state-linked capital two; US-based funds, Coatue and Addition, led the two largest rounds and most of the dollars; strategic and corporate investors are present, but the sample cannot say whether they lean toward participating or leading.
• Exits: thin in the 2025 regional data. On my read, the realisation pathway is the binding question for the cycle.

For the LP reading the late-stage SEA pitch deck in 2026, my read is that the due diligence questions that matter most are about the realisation pathway and the holding-period assumptions, rather than the pace of deployment or the quality of the next vintage of seed-stage candidates. If the 2025 exit data above is a guide, a growth fund underwriting a 2026 SEA late-stage commitment with 2018-vintage exit assumptions risks a mismatch with the cycle it’s deploying into.

If this sample is indicative, the reallocation has happened: the rounds in it look different from the consumer-platform rounds that defined the prior cycle, though without a sourced baseline for that cycle I can’t measure by how much. Whether the book being built now produces returns will depend on how well the realisation environment normalises over the next four to six years — and that is the question the sample cannot answer on its own.