On April 25, 2024, StackPath announced more computing capacity in Paris. Customers could provision virtual machines and containers with up to 48 virtual CPU cores and 256 GiB of memory. The pitch was pleasingly concrete: put the work near the people waiting for it. Forty-eight days later, the company stopped customer invoicing. In June, it announced that all products would close and its assets would be liquidated for creditors.
- The idea: run applications in metropolitan edge locations to shorten data’s journey.
- The backing: $396 million in reported equity funding by March 2020.
- The outcome: a CDN exit, a security-technology sale, then a June 2024 shutdown.
That juxtaposition is what makes StackPath worth studying. A company can announce a useful product while its future is running out. For anyone buying infrastructure, the question extends beyond how fast a request comes back. It includes whether tomorrow’s request will have somewhere to go.
The cloud has an address
StackPath began in 2015, led by Lance Crosby, the entrepreneur behind SoftLayer, which IBM acquired in 2013. At its public launch the following year, the new company described a fragmented security market: businesses were assembling separate appliances and services to defend the same internet traffic. StackPath wanted security built into the services themselves.
It bought capabilities rather than waiting to invent every component. MaxCDN supplied content delivery; Fireblade supplied web-application firewall technology; Cloak brought VPN technology. Highwinds joined in 2017, adding network reach and CDN expertise. In 2018, StackPath announced a consolidated platform combining delivery, security, DNS and monitoring. The proposition was fewer separate controls for the customer, backed by more machinery behind the curtain.
Its geographic argument became clearer as compute arrived. StackPath’s own materials described facilities inside major metropolitan markets, connected through a private network and controlled through one management system. A content delivery network brings copies of files closer to readers. Edge computing brings the program closer, too. The distinction matters when a response has to be calculated rather than fetched from a cache.
“The Internet has grown bigger and bigger but, from an end user or client perspective, it hasn’t grown closer and closer.”Lance Crosby, February 2019
One image, four cities
The February 2019 launch of containers and virtual machines made that proposition tangible. Developers could submit a workload image, choose its size and select deployment locations. This suited applications sensitive to delay: media delivery, security processing, advertising technology and connected-device services. StackPath’s attraction was familiar computing tools with geographic distribution available from the same controls.
The intended buyer might be a developer distributing an API, a media business moving content, or an enterprise deploying a virtual appliance. Company materials described customers ranging from one-person startups to Fortune 50 businesses. In 2020, Juniper Networks and Cox Communications were reported as customers as well as investors. Those overlapping roles suggested practical interest in the platform, though investment and customer adoption measure different things.
A useful contemporary experiment came from network engineer Andree Toonk. He deployed eight containers across Amsterdam, London, New York and Los Angeles, with an Anycast address. His small application printed its hostname so he could see which instance answered. The estimated bill was about $627 a month, including roughly $100 for Anycast. Geography had become a configurable resource, complete with a price tag.
The price of being nearby
StackPath charged for the infrastructure it supplied. Its 2018 Edge Delivery package started at $20 a month. At the 2019 compute launch, instances started at $0.091579 an hour. Later announcements described hourly instance charges plus transferred-data volume. These are historical prices, and distributing a workload multiplies the resources being rented. Toonk’s experiment shows why the cheapest individual instance is a poor proxy for the bill.
The comparison set depended on the job. Akamai, Cloudflare and Fastly occupied overlapping territory in delivery, security and edge services. AWS, Azure and Google Cloud offered broader cloud infrastructure. StackPath’s chosen distinction was proximity with general-purpose VMs and containers. A nearby machine offers less benefit if every request must consult a distant database. That is an architectural inference, and a reason to measure the entire request before buying the postcode.
Reported rounds backed by Abry Partners, then Juniper Networks and Cox Communications. Capital raised does not establish cash spent or creditor losses.
The retreat came in pieces
The withdrawal was gradual before it became abrupt. In August 2023, Akamai announced its purchase of approximately 100 enterprise customer contracts after StackPath decided to cease CDN operations. The announcement explicitly excluded StackPath personnel and technology. Buying customer contracts is quite different from buying the organisation that served them.
In March 2024, Gcore acquired StackPath’s web application and API protection technology. A mirrored StackPath status notice then recorded the retirement of CDN, WAF and serverless products, explaining that the company was concentrating on edge compute. Highwinds CDN retirement followed in April. This documents a narrowing strategy; it does not establish the private deliberations behind it.
Paris followed that narrowing. Then came June’s closure notice, urging customers to transition services and copy their data away. The sequence supports a precise account of what stopped and when. It does not support a confident diagnosis that one product, investor or acquisition caused the collapse. Nor should the funding total be casually relabelled money lost. Gcore’s subsequent WAAP launch shows that some technology found another home.
Copy the method, keep the exit
StackPath leaves a practical purchasing discipline. Package workloads so they can move. Automate deployments. Test response times from actual user locations. Compare the complete distributed bill, including traffic and dependencies. Keep backups outside the platform and rehearse migration. Edge placement earns its cost when distance materially delays the application. For a batch job with little urgency, or a service tethered to one remote database, extra locations may simply add expense. The cloud’s address matters. So does the route out.
Follow the trail
Explore the company’s historical presence and the transactions that followed.