In 2001 I wrote my LSE dissertation on why "ASPs" were failing. Here is the 25-year scorecard.

Fashion setters have been successful in preventing the ASP brand from acquiring a bad reputation as a failed innovation. This may give a newer and improved version of ASP the [chance] to succeed.

There are always innovators […] the first wave will take the hit, the second wave may take advantage of the lessons learned from the first ones.Anonymous senior consultant, global consulting firm, interviewed July 2001

The flagship pure-play ASP, USinternetworking, filed for Chapter 11 in January 2002; its shares had gone from $57.50 on IPO day in 1999 to $1 by mid-2001. Corio was absorbed by IBM in 2005. But the "newer and improved version" was already alive: Salesforce, founded in 1999 with the "No Software" logo, reported $41.5bn of revenue for fiscal 2026. The model is now simply how business software is sold.

The pure, first generation ASPs […] that have succeeded to date are those delivering services such as email, billing, online backup or automatic virus checking. These solutions are easy to implement and use, and one size can indeed fit all.

Hosted email became Gmail (2004), Google Apps (2006) and Office 365 (2011). Online backup and "automatic virus checking" became the cloud-delivered backup and security industries. Microsoft 365, the direct descendant of the hosted-email ASP, is now one of the largest subscription software businesses in the world.

Others have done well by focusing on tight vertical markets where they have specialist knowledge.

"Vertical SaaS" became a category of its own: life sciences, construction, restaurants, dental practices, each with a multi-billion-dollar cloud vendor. The most satisfying data point is personal. One of my 2001 interviewees was the deputy head of the technical department at Edicom, a Valencia company then providing EDI as a service. Edicom today describes itself as a SaaS platform for e-invoicing and tax compliance, with over 1,000 staff, 18,000 clients and projects in more than 85 countries.

It is easy to mistake the adoption attributes of the ASP concept with the attributes of the applications it delivers. […] The complexity in the ASP concept was very well encapsulated in the ASP acronym […] It is then a 'black box' of 'black boxes'.

The industry unbundled the black box into layers with their own names and their own economics: infrastructure (IaaS), platform (PaaS) and application (SaaS). Gartner now forecasts each separately; in 2025, $212bn, $209bn and $299bn respectively. The thing I struggled to name in 2001 because it was "mixed with the technology it delivers" is now three markets.

Classical models do not generally consider how an innovation could be changed, modified or adapted through more innovation over the original to make it conform to the needs of the real world. There is an underlying assumption that an innovation is a fixed non-evolving entity.

By using the terms ASPs 1.0 and 2.0 I am falling into the same mistake I am pointing to.

ASP became "on-demand", then "software as a service" (a term the Software & Information Industry Association had only just coined to tidy up the ASP/AIP/BSP/SSP alphabet soup), then "cloud" (2006 onwards), then "platform", and in 2026 is being repackaged again as "agentic". Nobody tracks the adoption of "the ASP innovation" any more because the thing being adopted kept changing its name and its shape, which is exactly what Figure 1 of the dissertation tried to draw. I also note, with some embarrassment, that I was using "2.0" as a version label for a concept three years before "Web 2.0" made it fashionable.

When dealing with the new forms of ASPs fashion setters had the choice of discarding the ASP 'brand' to distance the evolved forms of the concept from a failed first wave. Alternatively, it could have been made clear that something has changed and keep the existing brand. It seems that a compromise solution is being applied […] MAP (Managed Applications Providers) or MSP (Managed Service Provider) […] and some evolutions have adopted the term ASP 2.0.

The market chose the rename. "ASP" was quietly retired within two or three years and "SaaS" took its place; "MSP" survived, but for infrastructure management. The mechanism I described (the concept escapes the reputation of the companies by changing its name) is the one that operated.

ASPs still promise lower IT expenses as well as making it simpler to estimate a company's IT expenditure more accurately. This is very appealing in times where cost reductions are necessary. Therefore, ASPs and integrators believe that economic recession could be a catalytic start for people looking at ASPs in more detail.

The 2001 to 2003 downturn killed the venture-funded ASPs but did not kill demand. The 2008 to 2009 recession was the moment SaaS went mainstream in large enterprises: no capital expenditure, predictable subscription cost, and a way to cut the internal IT budget. "Opex, not capex" became the standard CFO argument.

It seems difficult to craft contracts for cross-national deals, which must consider different regional modes of operations, country laws and culture. This has limited the growth of ASP in the profitable large company market which has pan-continental or global operations.

This became data residency, GDPR (2018), the Schrems II ruling (2020) and the sovereign-cloud industry. Gartner expects sovereign cloud infrastructure spending alone to reach $80bn in 2026. Every hyperscaler now runs country-specific regions and legal entities precisely because of the problem in that paragraph. I thought it was a contracting nuisance; it turned into a geopolitical market.

Many of these large companies are hosting several applications themselves for internal use but also for their networks of partners and suppliers embracing some of the ASP ideas and competing with pure ASPs.

The largest cloud business in the world came out of a retailer. Amazon Web Services launched S3 and EC2 in 2006 by selling the infrastructure Amazon had built for itself and its marketplace sellers. Azure passed $100bn of revenue in Microsoft's fiscal 2026. The "water from the same reservoir" metaphor came true, but for computing rather than for applications, and it was delivered by companies that were not ASPs at all.

Some software vendors, including Oracle and Microsoft, still see the ASP model as their primary software distribution route in the future. […] IBM can afford to take this risk and have a larger trial and error experimentation process while the model is perfected than any start-up company.

Microsoft Cloud revenue was $59.3bn in the single quarter to June 2026. Oracle rebuilt its entire application suite for the cloud and bought NetSuite, one of the original 1998 ASPs, for $9.3bn in 2016. SAP now sells S/4HANA primarily as a cloud subscription. It took them a decade longer than the ASP believers expected, and it took several failed attempts each (Microsoft's original .NET "Hailstorm" services were cancelled in 2002), but the trial-and-error argument was the right one.

Another factor identified in this research is the power of investment banks in setting up and sometimes artificially maintaining fashions. Investment banks will sometimes use their position as opinion leaders to protect their investments in companies selling innovative solutions.

For twenty years the venture and public-market "SaaS playbook" (ARR, net revenue retention, the Rule of 40, cloud indices) was the most powerful fashion-setting machine in enterprise software. The same machine ran in reverse in early 2026, when a Jefferies trader coined "SaaSpocalypse" and a leading SaaS index fell 25.7% in the first quarter on fears that AI agents would replace per-seat software. Abrahamson's counter-bandwagon, which I cited in the literature review, arrived on schedule.

The role of professional organisations seems to be initially ignored by all of the interviewees in the ASP case and never mentioned when answering an open question about factors of adoption of ASPs. […] In the case of ASPs professional organisations are eclipsed by the media.

The ASP Industry Consortium, founded by 25 technology companies in May 1999, was folded into CompTIA in 2001 and has left almost no trace. What actually created the confidence and de facto standards I said professional bodies could have provided came from elsewhere: analyst rankings, open-source communities, developer conventions (REST APIs, OAuth) and audit regimes (SOC 2, ISO 27001).

The original ASP model of one-to-many solutions has proven flawed due to the higher than expected levels of customisation that the client required and demanded. ASPs were not prepared for this, and the technical infrastructure and software packages required to provide this customisation were not available. […] ASPs are currently changing their original model […] moving from the multi-user idea to a multi-customer approach.

The diagnosis was right: first-wave ASPs were hosting single-tenant copies of client-server packages such as SAP, Siebel and Oracle 11i, and every customisation was a separate server to maintain. But the cure my interviewees were describing (retreat from one-to-many towards bespoke hosting) was the wrong direction. The winners doubled down on one-to-many: a single multi-tenant code base where every customer's "customisation" is configuration and metadata, plus APIs and an app marketplace (Salesforce's AppExchange, 2005) for everything else. One size did fit all, once the software was rebuilt so that it could.

The service level agreement (SLA) has been identified in this research as the major social problem and the major overall problem in the diffusion of the ASP innovation. […] It seems unclear at this stage of ASP development if a good model of SLA that provides a win-win situation for both the client and the service provider can indeed be created.

It was a real blocker in 2001 and it did get solved, but not the way I recommended. Following the outsourcing literature, I argued SLAs should be tailored to each client. The market went the opposite way: standard, published, take-it-or-leave-it terms (99.9% availability, service credits, a public status page, a SOC 2 report) that the customer accepts by clicking. The "political problem of mixed and opposed interests" was dissolved by standardisation rather than negotiated away. Only the very largest customers still negotiate, and even they start from the vendor's paper.

It has proven very difficult to make enough money from selling simple applications to SMEs. Larger clients would be needed to make these services profitable and achieve important economies of scale. […] ASPs that cannot penetrate large companies will have difficulties to survive.

Half of this is exactly what happened: Salesforce, Workday and ServiceNow all became large by moving upmarket into the Fortune 500. The other half is wrong. The SME long tail became one of the most profitable segments in software once the cost of selling to it collapsed to almost nothing: self-service sign-up, free tiers, product-led growth. Shopify, Xero, Intuit and Zoho are businesses built on the customers my interviewees said could not be served profitably. The assumption that failed was not about SMEs' willingness to pay but about the cost of a sales call.

The most important effect will be that software providers and consultants with outsourcing capabilities will have an even bigger role in the development of ASPs. […] Small companies cannot offer this kind of experiment particularly after the dotcom burst and the scarce venture capitalist money to unproven businesses.

The incumbents did take a very large share of the eventual market, and the big consultancies became the implementation partners of every major SaaS vendor rather than its victims. But the category was defined by start-ups: Salesforce, NetSuite, WebEx, Concur, Workday, ServiceNow, Shopify, Zoom. Venture money did not stay scarce for long. I over-weighted the network advantage of large firms and under-weighted how quickly a start-up with a better architecture could build its own network.

Microsoft's .NET initiative regardless of whether it is successful or not, will take some of the ASP ideas to the masses. […] The failure or success of the .NET initiative will provide some first hand exposure to ASP related concepts to decision makers (through home use for example) and this may influence their opinions when taking corporate decisions.

.NET as a consumer web-services platform failed (the "Hailstorm" services were dropped in 2002). But the mechanism was right: decision makers learned to trust software-over-the-network at home first, through Hotmail, Gmail, iTunes and later Dropbox, and carried that comfort into the office. Microsoft eventually delivered the vision through Office 365 (2011), a decade later than the sentence implied, and today counts over 30 million paid Microsoft 365 Copilot seats.

Traditional IT outsourcing and in principle ASPs do not generally carry any connotations of prestige. A company will not gain or lose prestige for outsourcing its IT operation. […] This seems to be an important limiting factor in the adoption of ASPs by imitation forces.

Within a decade, "cloud-first" was a badge that CIOs put on conference slides and boards put in annual reports. Netflix's public migration to AWS between 2008 and 2016 did for cloud what I said Kodak had done for outsourcing in 1989: it "authenticated the marketplace". Salesforce's entire early marketing was built on making the alternative look old-fashioned. Imitation and prestige turned out to be among the strongest forces in the whole story, and I had written them off.

Technical problems with data conversion and data transfer, problems with more or less serious security glitches and software bugs are all sources of complexity. However, this research clearly identifies the three main sources of initially ignored complexities in ASPs as being the most important problems affecting the diffusion of the model.

The three problems I named (customisation, contracts, revenue) were real, but the things that actually unlocked the market were technical and I barely mentioned them: broadband replacing dial-up and leased lines, the browser becoming a real application interface (AJAX, 2005), multi-tenant metadata-driven architectures, virtualisation and commodity data centres. In 2001 "bandwidth" appeared once, in my interview questionnaire. Seen from an information-systems department, everything looked like a social phenomenon. Some of it was engineering.

ASPs are more of an evolution of IT outsourcing rather than a revolution, [which] gives this innovation a smaller public profile.

This was the consensus in 2001 and it was the framing mistake of the whole industry, including me. ASPs were bolted onto the outsourcing story (Kodak, EDS, ten-year contracts, a retained IT department). What SaaS turned out to be was a change in how software is built, priced and bought, not a change in who runs the servers. The lineage that mattered was not Kodak-to-IBM in 1989; it was the web.

The dissertation treats price as a detail of the SLA. In fact the subscription per seat per month, later joined by free tiers and usage-based pricing, was the mechanism that changed the buyer (from CIO to department head to individual user), the sales motion (from a relationship to a credit card) and the vendor's economics (from a licence sale to a lifetime value). The 2026 argument about whether AI agents break per-seat pricing is an argument about this, and I had no vocabulary for it in 2001.

I saw ASPs as a channel for someone else's software. Nobody I interviewed imagined that the hosted application would expose APIs, run third-party code and host a marketplace, and that the ecosystem would be the moat. Salesforce launched AppExchange in 2005 and its platform in 2007; by then "ASP" was a historical term.

The "water from the reservoir" metaphor was about applications. The reservoir turned out to be computing itself. Every SaaS company after about 2008 was built on somebody else's rented infrastructure, which removed the capital cost that had killed the first-wave ASPs, who had all built their own data centres. Gartner's 2025 figure for IaaS alone is $212bn.

My "user-focused perspective" meant managers actively searching for innovations. The actual users turned out to be employees signing up for Dropbox, Slack and Zoom with a personal email address and presenting IT with a fait accompli. Bottom-up adoption, not the boundary-spanning manager, was the diffusion channel that Rogers' framework and my interviewees both missed.