Hey! Thanks for being here for Edition #18. You're one of 315+ subscribers, and I genuinely appreciate every single one of you.
I accepted 450 new connections this week, so if you're new here, welcome. A quick bit of context on who's writing this.
I'm 35, born in the UK, raised in Australia, and I've lived in seven countries. These days I'm based in Saigon. My first jobs were as a delivery driver and a retail rep. From there, I moved into market intelligence for heavy industries, did a master's in cybersecurity, worked in sales, and built a B2B advertising and video agency for IT firms.
Then I realised I hated what I'd built.
So I burnt it down and started from zero in private markets, speaking to anyone who'd take my call. Now I work in an industry I love: putting the right people in the same room, then making sure the deal gets done.
This month I'm focused mainly on ASEAN and APAC, which fits nicely with this week's Vietnam story. And one theme ran through almost everything I saw this week: capital isn't the bottleneck in data centres right now. Power is.
What should I cover in Edition #19? Hit reply and tell me.
Let's get into it.
Singapore's GIC is selling $1B of private equity - one of the biggest LP secondary deals Asia has seen. HarbourVest Partners is buying the package of fund stakes, mostly Asian PE exposure. GIC doesn't need the cash. It's reshaping its PE book, and secondaries are the fastest way to do that.
It isn't alone. The Teachers' Retirement System of the State of Illinois has just put out an RFP for PE secondaries advisers, due late October, with a plan that spells it out: sell stakes to generate distributions, concentrate on high-conviction managers, and manage cash-flow pressure. PE distributions have slowed, but LPs still have commitments to fund and members to pay - so they're making their own liquidity. That's pushing more deal flow into LP stake sales, continuation vehicles and discounted interests, and when sellers line up, buyers set the price. Selling on the secondaries market used to look like a sign of trouble. Now sovereign wealth funds and pensions do it on purpose.
Is the secondaries market becoming PE's real exit?
Also this week:
A $15B AI campus outside El Paso - backed by a family office, not a hyperscaler. Kaufman & Company has broken ground on a 350MW campus on a 24,000-acre former ranch, backed by The Kaufman Family Office (investing since 1956). That works out at roughly $43M per MW. Up to 90% of the power will be generated on-site from geothermal and solar, outside ERCOT, with the first 80MW due 16 months after construction starts. The build partner is ServerDomes, a Boise dome-data-centre designer Kaufman Ventures has also invested in. Phase one tenants are claimed but not named. Family money, no fund deadline, no pressure to sell.
Vietnam: $1.45B across four deals, and three are AI hardware. AVC ($500M, 2027 plan) makes heat sinks and liquid cooling inside NVIDIA's supply chain, with $760M already in Ninh Binh and 14,000 staff. Sembcorp's $480M JV is building its first Vietnam data centre in HCMC (Starmason, 60MW expandable to 90MW, land lease to 2076), while G-Group's $300M Hoa Lac campus starts at 20MW liquid-cooled. Excelsior Capital Vietnam closed a $170M second fund - every Fund I LP re-upped - writing $8-25M tickets into the middle-class economy. The risk is power: the northern grid buckled as recently as 2023.
SpaceX pays ~$8B to become a phone carrier. It's buying Grain Management's 800 MHz spectrum, all cash, pending FCC approval - the low-band piece Starlink Mobile was missing. AT&T, Verizon and T-Mobile sold off; Crown Castle, American Tower and SBA rose. The private-markets angle: Grain bought that spectrum from T-Mobile in August for $2.9B plus its 600 MHz licences. Two months later it's selling for $8B. SpaceX is also reportedly lining up $40B of debt for Nvidia chips, with Apollo expected to lead.
What's crossing my desk and my conversations this week:
A US balance-sheet lender offering up to 100% financing. Credit committee in the same building, $1M to $100M+ per deal, covering labour, build-out, demolition, renovation, ERP and software rollouts, and ground-up projects. Strong credit prices around prime + 2-3% on 2-5 year terms. The catch: 3+ years trading with historical financials, $15M+ revenue, US and Canada only, and no start-ups, cannabis, gaming or inventory finance. They just closed $25M on 32 nodes of NVIDIA B300s and have $50M to deploy before November. He says no to brand-new data centre developers. He told me he gets calls every day from data centre developers asking for $50-75M, and some of these companies are only four months old. He turns them all down, no matter what offtake they have. As he put it: "100% only works if you've already proven you can run the business."
Would you take 100% debt if it was offered?
135MW powered-land site, Hungary
This one is a sale of powered land, not a capital raise. 163 hectares of industrial-zoned land in an EU member state, with 45MW of grid power ready to connect now and a further 90MW at 132 kV, earliest Q4 2029. Two separate dark-fibre routes run toward EU backbone markets, and a dedicated raw-water source has been identified for cooling. The buyer takes the land and grid capacity, then funds the build. Worth noting: the teaser's 300MW+ headline relies on behind-the-meter power that doesn't exist yet, and there's no price, capex budget or tenant disclosed.
It's pre-development - no buildings and no data centre permit yet - though the UXO survey, geotech and access permits are done. Municipal support is an LOI, not an approval. Downside protection comes from the 45MW of real, documented capacity and zoning already in force. The weak spot is that the extra 90MW is an offer, not a contract, and the buyer pays for the grid upgrade.
That matters for the exit. The route out is a hyperscaler lease or a sale, but no tenant interest has been disclosed, and a 30-35MW Phase 1 is small for a hyperscaler. The upgrade is what truly makes this a hyperscale site.
My verdict:
Strong fit for hyperscalers
Strong fit for data centre developers
Medium fit for infrastructure funds
Weak fit for private credit
Why I don't sign NDAs before I see a teaser
If you need finance, show the deal first.
When I sign an NDA up front, my clients then have to sign one too - two rounds of paperwork before anyone knows whether the deal even fits. That friction slows everything down and kills momentum.
What works better is simple. Send a no-names teaser, summary or deck: asset, location, size, the ask. Just enough for someone to say yes, no, or tell me more. If there's interest, we sign the NDA, you open the data room and share the full details, and I can make the right introductions quickly.
Would you sign an NDA before seeing a deal?
No new deals since last week, so I'm using this space for something else.
What do you want to see more of in Capital Arbitrage Weekly? More deal breakdowns, more lender conversations, more market intel, or something I'm not covering yet?
Hit reply and tell me. I read every one.
That's all for this week. See you next Friday.
— Jordon
P.S. I get 500,000+ monthly impressions on LinkedIn and a growing list of private capital readers right here. If you'd like to get your company, fund, or raise in front of this audience, just reply to this email.