By Viktor Yoshimura – 29th October 2026
Scenario. A composite built from common situations, not a client engagement.
The short version
Partner with one licensed Japanese processor now, run it to our standard, and prove it in a pilot before we commit. Review a minority stake, with governance rights, after 12 to 18 months of clean audits. Building takes longer than the client will wait, and buying cold means proving ourselves to a seller from a standing start.
The situation
A European ITAD provider serves a US hyperscaler in Singapore and Australia. The client is bringing two data halls online in the Tokyo area and wants the same provider in Japan within twelve months, starting at around 3,000 to 5,000 servers a year and rising. The provider has no entity, no licences and no people in Japan. The regional MD wants a recommendation by the end of the quarter: build, buy or partner.
What’s actually at stake
The stated problem is capability. The real problem is time and trust. Twelve months is short in Japan for anything that needs a licence, and the hyperscaler’s facilities team will audit whoever turns up on site, whatever logo is on the contract. Lose the Japan sites and the provider becomes a two-market supplier to a client that’s consolidating vendors across the region. Get it wrong in front of a Japanese partner and the damage outlasts the contract.
Three constraints sit underneath. Collecting, transporting and treating industrial waste each need permits under the Waste Management Act, issued by the prefecture or designated city where the activity happens, and specified treatment facilities need a separate installation permit, often with local consultation in practice. Dealing in used equipment generally needs a secondhand dealer licence. Which of these the provider needs depends on what it does itself and what a licensed partner does, and that should be confirmed with a Japanese compliance specialist before anything is signed. Separately, Japanese clients often expect witnessed destruction and paperwork that matches the manifest line by line.
Before deciding anything, I’d want four things on paper:
- The client’s volume forecast and contract term. A 3,000-unit year one means little without years two to five.
- The destruction specification: on site or off, witnessed or not, certificate turnaround.
- What the client will accept from a subcontractor, and what it expects us to own directly.
- Which licensed processors in Kanto already handle data centre work, and how full their floors are.
And three tests sit above the ownership question. Will the client’s security and supplier audit accept the structure? Who is accountable for asset-level reconciliation, sanitisation evidence, chain of custody and downstream disposition? And what happens to us if the forecast volume doesn’t arrive, or arrives late?
The options
| Option | Upside | Risk | Time to first job (this scenario) | Reversible? |
|---|---|---|---|---|
| Build: own entity, site and licences | Full control and margin | Licences and local consultation take longer than the client will wait; no track record to show the auditor | 18 to 24 months | Costly to unwind |
| Buy: acquire a local ITAD | Licences, staff and references from day one, if it’s a share purchase and permits and contracts survive the change of control | Finding a willing seller; an asset purchase may need new permits; integration and staff retention; paying for a business you don’t yet understand | 9 to 15 months | Hard |
| Partner: subcontract to a licensed processor | Fast; uses existing licences and audit history | Shared margin; your SLA depends on their floor; you’re one customer among many | 3 to 6 months | Easy |
| Partner now, option to invest later | Partnership speed, with a path to control once trust exists | Needs a partner open to discussing equity; governance has to be designed in from day one | 3 to 6 months | Mostly |
Table 1. Scenario assumptions, not benchmarks. Greenfield includes site, permits, staff and systems; partner-led depends on client approval and the partner’s existing licences.
The call, and why
Partner now, with a structured path to a minority stake. Pick one licensed processor in the Kanto region with data centre experience, run the hyperscaler work through them under our process and reporting, and agree up front that both sides will review a minority investment after 12 to 18 months of clean audits. Governance rights get written in from the start, because a minority stake on its own doesn’t give operational control.
For this scenario, build misses the clock. Buy is possible but slow and uncertain. An owner-managed business, in Japan as anywhere, wants to know a buyer will protect its staff, customers and service, and a foreign buyer with no local track record has to prove that from a standing start. Plenty of owners of mid-sized firms are thinking about succession; the buyers they trust tend to be the ones they already know. A partnership is the lower-risk way to become one of them.
The client is buying reliability in Japan, not our ownership structure.
The numbers
Rough, and every figure is an assumption, but enough to see the shape. A build carries a site, fit-out, permits and a team of six to eight before the first server arrives. Call the running cost ¥60 to 90 million a year. At 3,000 to 5,000 servers that’s ¥12,000 to ¥30,000 of fixed cost per server before anyone touches one. A partnership costs little up front, perhaps ¥5 to 10 million for audits, onboarding and systems integration, and gives the partner a share of the gross profit on every job. Say a third.
Fixed cost per server isn’t the break-even, though. Building pays when the margin we’d otherwise hand the partner, per server, times annual volume, covers the build’s running cost plus a return on the money put in. If the partner’s share works out at ¥10,000 a server, ¥60 to 90 million of running cost needs 6,000 to 9,000 servers a year before the investment is repaid. That’s above year one, and only reachable if the forecast holds.
Change the assumptions and the threshold moves, which is the point: it should come out of the model, not go into it. At this volume and with this much uncertainty, the partnership is the cheaper and safer way in.
30, 60, 90 days
| Actions | How we’ll know it’s working | |
|---|---|---|
| Day 30 | Confirm volumes, spec and audit criteria with the client in writing. Long-list licensed processors and confirm the permit position with a local specialist. Visit the top three with a Japanese-speaking colleague, walk the floor, follow one manifest from dock to certificate. | Client sign-off on the spec. Two viable partners, not one. |
| Day 60 | Choose one partner. Run a readiness pilot on a limited scope: our reporting template, security controls and governance tested on a small test batch. Write the governance in both languages: joint KPIs, audit rights, data handling, escalation. | Every test asset reconciled from intake to certificate. Reports in our format, on time. |
| Day 90 | Client audit of the partner’s floor and approval of the structure. Agree commercial terms and a monthly operating review. Schedule the first live job. | Zero major audit findings. Client approval in writing. First live job booked for month four to six. |
Table 2. Owners: account lead (client), regional operations (long-list), the author (visits, pilot, governance). The live operational pilot follows client approval.
What would change my mind
- The client commits to volume clearly above the break-even in The numbers, over several years. Then I’d rerun the model and start a build alongside the partnership.
- A suitable owner approaches us about succession during the first year. Then buying stops being a cold acquisition and becomes a known relationship.
- The partner fails a client audit or can’t produce asset-level reporting. Then we change partners before anyone talks about equity.
The lesson
Where trust is the scarce asset, buy time with a partnership before you buy a company. Ownership can come later. Trust can’t be acquired on day one.
How would you call it?