Flag's second act: selling the AI on-ramp

Subsea cable operators usually get one shot. The capital is too heavy, the demand too cyclical, and the physical assets too hard to duplicate, so the industry is full of networks that were built once, under one owner, and retired with that owner's excuses. Flag has managed a second shot, and the company now in its fourth year under the ownership of London investment firm 3i Infrastructure is spending it on the biggest reason any cable operator gets a second chance today: connectivity to AI data centers.

Data Center Dynamics retells the sequence in a profile published this week: the company, then known as Global Cloud Xchange, filed for bankruptcy in 2019 with $350 million in debt, reemerged in 2021, and was acquired by 3i Infrastructure in September 2022 for $512 million. Four years on, the operator has dropped the GCX initials, restored the Flag name, and repositioned itself for the AI era.

Carl Grivner, Flag's chief executive, has been with the company for six years as of July, he told DCD; he arrived as the company was emerging from bankruptcy and describes the intervening years in plainly commercial terms. "Since then, we reshaped it, rebuilt it, and then we sold it to a private equity firm here in London," he says.

The reshaping included a rebrand announced in April of last year, and Grivner says the old moniker was a burden, especially with the over-the-top content players Flag now courts. "It's quite comical," he says. "The original name was Global Cloud Xchange, which was then translated into GCX... a lot of the OTT players hated it." He remembers the reception the name got: "It was a 'wait a minute, you're not in the cloud business and not in an exchange in any way, so why are you called GCX? It was a painful discussion to have."

The Flag name was more than an available trademark; it was an asset. The company's cable systems share the name — FLAG Europe-Asia, FLAG-Atlantic and FLAG North Asia Loop — and Grivner says it still draws affection inside the industry. "We decided to go back to Flag, which was a historical and very successful brand in the industry," he says. "When you're at events, it seems like everybody worked at Flag at one point or another. People cherish the little memorabilia, such as the pins you put on your jacket, or the silly socks that say Flag and things like that."

Grivner's resume explains why the brand mattered to him: he spent four and a half years as CEO of Colt Technology Services and previously served as chief executive of Singapore-based Pacnet, two carriers where the commercial game was winning enterprise and wholesale accounts on the strength of relationships. That is the game Flag is trying to play, only now the counterparties are the hyperscalers and content platforms spending on AI.

The asset underneath the brand is substantial. DCD describes Flag as owning one of the world's largest private subsea fiber networks, with 66,000 km of cable stretching from North America to Asia; the network includes seven subsea and six terrestrial cable systems, capacity on the FA-1, HAWK, Falcon, MEDEX, FEA, IEX and FNAL systems, and the company operates or arranges capacity at cable landing stations in Taiwan, Japan, France, the UK and the US.

Flag also sells ground-side product: colocation capacity, dark fiber and Layer 2 and 3 services for hyperscalers, telcos, OTT content providers and enterprises, plus modular data center services. Add the pieces together and Flag has become a builder and operator of the physical layer on both sides of the beach.

Put the bankruptcy next to the acquisition and the strategy comes into focus. GCX went down in 2019 carrying $350 million in debt; 3i Infrastructure paid $512 million for the reorganized business in 2022, a price that was less a bet on the old company deciding to be better than on the assets — 66,000 km of fiber, landing stations in five jurisdictions, and a route map from North America to Asia — being worth more under a different owner.

The change in ownership is also a change in financial regime: under GCX's old structure, the network was in service of a telco's broader ambitions, while under 3i, Flag is an infrastructure asset with a private equity owner's return expectations — contracted revenue, predictable maintenance, and a book of business long enough to support both the cable and the landing stations. That is a different discipline from the one that left the company in bankruptcy.

The AI buildout is why that discipline is showing a payoff: the data center boom is usually narrated as a power story, and it is one, but every gigawatt of AI compute needs an on-ramp to the network, and the on-ramp runs through cables and through the buildings at the shore where those cables land. Flag's modular data center services put it at the junction where the cable ends and the data center begins.

This publication has argued that grid access has become the scarce asset class in digital infrastructure, more binding than the data center shell itself. Landing stations are the coastal version of that argument. A cable can be rerouted; a landing station is a point of physical presence with permitting, power and real estate attached. Flag has those points in Taiwan, Japan, France, the UK and the US.

The risk is on the offtake side: Flag's pitch works only if hyperscalers and large content companies commit to capacity and colocation at the landing stations, rather than buying bandwidth as a spot product. Subsea cable history offers enough examples of systems built on projections that ended up selling merchant capacity. Flag's ownership structure does not protect it from that outcome; it only raises the cost of getting it wrong.

Grivner's record suggests he understands the distinction: both Colt and Pacnet operated in markets where customers could have built around them, which is the same environment Flag now confronts with hyperscalers. The question is whether Flag can sell colocation, dark fiber and Layer 2 and 3 services as one integrated answer to a hyperscaler's problem, rather than as a list of line items. The integrated answer justifies the modular data center investments; the line-item version is how networks end up as commodity bandwidth.

The ownership structure points the same way. Infrastructure investors do not pay $512 million for an asset and then leave it to sell bandwidth by the gigabit. They build a base of contracted tenants, and the asset that anchors those contracts is the colocation and data center capacity at the beach. Flag's modular data center offering is the tell: the company is trying to move up the capital hierarchy, from carrier selling routes to infrastructure owner selling connections into the AI buildout.

No one will mistake Flag for a global cable giant. But scale was never the point of this revival. The point was to take a network broken by its operator and repackage it as infrastructure for an owner that values long-lived assets. The 66,000 km of fiber is the proof of concept. The landing stations and the modular data centers are the thesis. Whether that thesis gets underwritten is now a question of who signs the first big ground-side contract.

The rebrand was announced in April of last year, and in the year since the AI data center conversation has only gotten louder while Flag has had a clear name to put behind its pitch. The pins and socks are back on jackets. The memorabilia that will matter is the anchor tenant agreement at a landing station in Japan or Taiwan, where the cable comes ashore and the AI era begins.