The Life & Services Supply Strength Index is one composite score out of 100 built from four weights, and the weights are the argument. Nothing in the formula pays for a famous name, a large office or a press release.
What Each Weight Measures. 40% goes to global revenue and absolute market share, which is why the largest balance sheets start near the top. 30% goes to target-category service revenue and business purity, counted only after hardware manufacturing and first-party retail goods value are removed from the base. 20% goes to owned infrastructure and physical assets: data centres, subsea cable, aircraft, warehouses and delivery networks. 10% goes to customer scale and global fulfilment coverage, read through client counts, net dollar retention, remaining performance obligations, sovereign countries served and headcount.
How The Ten Land. Microsoft scores 98 on 80.5 percent service purity, US$281.72 billion of revenue and US$678 billion of commercial RPO. Amazon scores 98 on the largest revenue on the page, US$716.92 billion, held back by a 58.7 percent service share because a first-party goods book sits inside the same accounts. Alphabet scores 98 on 91.8 percent purity and a US$242.8 billion contract backlog. Accenture scores 97 on 100 percent purity, US$69.67 billion and 779,000 professionals. DHL scores 96 on 86.9 percent category revenue and 295-plus owned freighters. Tencent scores 94 on 49.8 percent purity and RMB 374.4 billion of in-category services. Alibaba scores 94 on more than 60 percent purity and four million paying cloud customers. Oracle scores 93 on 95.3 percent purity and US$638 billion of RPO. Salesforce scores 91 on 100 percent purity and US$72.40 billion of RPO. SAP scores 90 on 100 percent purity and €36.8 billion of revenue, the smallest total here and the closest to the line.
Why The Range Is Nine Points. All ten clear the US$32.2 billion entry line of the 2025 Fortune Global 500 on their own reported revenue, from Amazon at US$716.9 billion to SAP at about US$40 billion. A band that no longer separates qualifiers from non-qualifiers has to separate peers, so the whole table is compressed into 90 to 98 and the order inside it turns on purity, owned assets and contracted backlog rather than on size alone.
What The Score Is Not. It is not a share price target, a credit rating or a forecast. It reads disclosed figures for one fiscal year and applies fixed weights to them, and a company that changes its mix changes its score. Two firms can sit one point apart and run very different businesses, which is the intended result of a narrow band.
Disclaimer: Figures come from published accounts and exchange filings for the fiscal years named, and revenue, backlog and headcount are given in the units each company reports. No figure on this page has been estimated on a service provider's behalf, and where a company discloses nothing for an item, the entry says so.
Remaining performance obligations and net dollar retention are the closest thing this page has to a forward revenue line, and the ten disclose them in four different shapes. The figure is only useful once a reader knows what sits inside it.
What The Backlogs Actually Hold. Microsoft reports commercial remaining performance obligations of US$678 billion, which is contracted work sold and not yet delivered. Oracle reports US$638 billion, more than four times its prior-year level, and over US$75 billion of that is customer prepayments and custom AI supercomputer contracts. The AWS backlog at Amazon is US$244 billion and the unfulfilled contract backlog at Alphabet is US$242.8 billion. Salesforce holds US$72.40 billion of total RPO, 14 percent above the prior year, against revenue growth of 10.0 percent. SAP describes a cloud backlog in the tens of billions of euros with recurring revenue near 85 percent of the total. Accenture signed US$80.62 billion of new bookings, US$37.64 billion in consulting and US$42.98 billion in managed services.
Why Duration Matters More Than Size. Oracle's backlog is the largest relative to its own revenue, US$638 billion against US$67.40 billion, and that gap is a construction schedule as much as a sales figure: the company carried a free cash outflow of US$23.7 billion to build the capacity those contracts require. Salesforce's backlog is smaller but shorter and more predictable, and its 34.1 percent non-GAAP operating margin shows what a renewal base does for unit economics. A backlog that must be delivered with heavy new capital is a different asset from one delivered on infrastructure already paid for. Microsoft's US$678 billion is the largest in absolute terms and is weighted toward multi-year Azure commitments, while Alphabet's US$242.8 billion sits beside a cloud run rate above US$70 billion.
Where Net Dollar Retention Fits. Net dollar retention measures whether existing customers spend more each year, and only a few of the ten publish it in a form a reader can compare. Where the number is missing, the page reads contracted backlog, customer counts and renewal terms instead: DHL serves more than three million business customers who buy multi-year contract logistics, and EBIT in that division rose 24 percent; Tencent connects more than ten million organisations through WeCom; Alibaba Cloud bills more than four million paying business customers; SAP counts more than 253,500 enterprises.
Three Questions Before Trusting A Backlog. Is it contracted or merely forecast? Over what period does it convert to revenue? And does converting it require capital the company has not yet spent? Salesforce's US$72.40 billion, Oracle's US$638 billion and Accenture's US$80.62 billion of bookings are all real, and none of them means the same thing. A backlog is a promise with a schedule attached, and the schedule is the part a reader can check.
Oracle installed its own Exadata hardware inside the data halls of Amazon Web Services, Microsoft Azure and Google Cloud, with 23 deployments live and 47 under construction, and that is the clearest test on this page of whether multi-cloud reduces lock-in or simply relocates it.
What Multi-Cloud Looks Like Here. Salesforce runs its workloads through Hyperforce, a control layer built in house that places them in Amazon and Microsoft facilities across regions. SAP runs SAP Converged Cloud, a company-operated network of 70-plus certified data centres, and opens sovereign-certified hosted zones inside AWS, Azure and Google Cloud. Oracle keeps 29 Cloud@Customer sites on customer premises where the data never leaves the building. Alphabet operates 40 cloud regions with 121 availability zones, Alibaba Cloud 30 regions with 89 availability zones and Tencent more than 30 regions, so a large buyer usually has at least three credible venues for the same workload.
Where The Lock Actually Sits. Lock-in has moved off the machine and onto three things a buyer signs: data egress pricing, licence terms that treat a rival's cloud as a penalty, and the cost of retraining the people who operate the stack. The European Commission and the British competition authority are examining the bundling of Teams into Microsoft 365 and the licensing terms attached to Azure, and that inquiry is about exactly this question. The Department of Justice liability ruling in the search distribution case against Alphabet concerns a different kind of default, but the same principle holds: whoever sets the default sets the price.
Why Rivals Still Share A Building. Oracle's database runs inside competitor facilities because customers asked for it, and the line has produced triple-digit sequential growth. Microsoft and Alphabet both sell capacity to firms that also compete with them. The commercial logic is that a shared venue expands the market faster than exclusivity protects it, and the buyer gains a negotiation position rather than a guarantee. The same reasoning explains why Salesforce places workloads in two hyperscalers rather than one, and why SAP opened sovereign-certified zones inside three rival clouds instead of expanding its own estate.
How A Buyer Should Read It. Multi-cloud reduces the cost of leaving and raises the cost of operating, because running the same workload in three venues needs three sets of skills and three sets of controls. The test is not how many clouds a company uses but whether its data can be moved without renegotiating a licence. On that test Oracle's US$638 billion of contracted work, Salesforce's US$72.40 billion of RPO and SAP's cloud backlog are all bets that customers will stay for reasons other than captivity. Egress charges are the line item that usually decides it, because moving a petabyte out of one venue can cost more than a year of running it there.
Accenture and DHL are the only two firms on this page whose main asset leaves the building each evening, and the financial shape of that model is visible in every margin they report.
The Two Labour Businesses. Accenture employs about 779,000 professionals and delivers through more than fifty advanced technology delivery centres and twenty-plus innovation studios across 120-plus countries. It reported US$69.67 billion of revenue for the year ended 31 August 2025, an adjusted operating margin of 15.6 percent, free cash flow of US$10.90 billion and record new bookings of US$80.62 billion. DHL employs about 594,000 people, more than 116,000 of them certified international specialists in Express, and earned EBIT above €6.2 billion on €82.9 billion of revenue, a margin near 7.5 percent. Both models pay for people before they invoice for work. Neither company owns most of the buildings its people work in, and neither can book revenue before the work is done, which puts payroll and working capital on the same side of the ledger. Headcount is the capacity constraint in both cases.
The Eight That Sell Capacity Instead. Salesforce turned US$41.50 billion of subscription revenue into a 34.1 percent non-GAAP operating margin with about 73,500 staff and free cash flow of US$14.40 billion. Microsoft reported US$101.83 billion of net income on US$281.72 billion of revenue with 228,000 employees. SAP lifted IFRS operating profit to €9.60 billion on €36.8 billion, and Oracle produced a record US$32.00 billion of operating cash flow. Those margins come from software and capacity that can be sold again without being rebuilt, which is precisely what an hour of consulting cannot do.
The Lever That Closes The Gap. Accenture's answer is the offshore delivery mix: an hour of consulting sold in New York or Frankfurt can be built in Bangalore, Manila, Poland or Dalian, and that spread is how a labour business sustains 15.6 percent while paying developed-market salaries to the partners who win the work. DHL's answer is network density, because 295-plus owned freighters, three intercontinental hubs and more than 115,000 service points let it sell the same lane to many shippers. Neither firm can raise margin by shipping a new version.
What Changes Next. The labour model is the one most exposed to automation, and the ten answer it differently. Accenture booked US$5.90 billion of generative AI work in the year, nearly double the prior year, and bought the British AI firm Faculty in March 2026 to add algorithm governance. It also spent about US$865 million on a six-month workforce and business optimisation. Salesforce moved pricing from seats to resolved tasks, with 29,000-plus Agentforce agreements and 2.4 billion agentic work units delivered. The question for both models is the same: whether the work being billed survives being automated, and who captures the saving.
Replacement cycles on this page run from a weekend to a decade, and that difference explains why some scores rest on growth while others rest on inertia.
The Longest Cycle. SAP has set 2027 as the end of mainstream support for legacy on-premise ERP and is pressing customers toward S/4HANA Cloud. More than 253,500 enterprises run its systems, and roughly 87 percent of physical trade flows and production chains depend on them, so the deadline converts patience into a schedule. Cloud revenue passed €21.00 billion, up 26 percent, and exceeded on-premise licence support for the first time. The friction is documented: German and European manufacturers have objected publicly to upgrade effort and cost, and SAP has restructured with voluntary departure and reskilling programmes touching more than ten thousand staff. It also paid US$1.5 billion for WalkMe to automate cross-system user flows, which lowers the cost of the move without removing the deadline.
The Database Cycle. Oracle serves more than 430,000 enterprise and government customers, many of them running core banking systems and general ledgers. Replacing a system of record takes years, which is why 29 Cloud@Customer sites exist for buyers who will not move data off their own premises at all. Oracle's own numbers show what the installed base is worth: US$638 billion of remaining performance obligations and cloud and support revenue at 95.3 percent of the total.
The Short Cycle. Adding a seat is cheap, replacing a platform is not. Microsoft holds more than 420 million Microsoft 365 commercial seats and more than 320,000 organisations paying for Copilot, so expansion arrives inside tools customers already open. Salesforce holds a 21.7 percent CRM share for an eleventh consecutive year across more than 150,000 enterprises, and the fulfilment network at Amazon carries more than two million third-party sellers who would have to rebuild logistics to leave. Tencent's 4.3 million active mini programs and 1.418 billion monthly accounts work the same way: the switching cost is the ecosystem, not the licence.
How To Price The Cost Of Leaving. Three figures settle it for any buyer. What fraction of the workload can be moved without renegotiating terms? How many years of retraining does the move require? And who holds the data model that everything else depends on? SAP's 2027 date answers the first question for its own customers by removing the option of staying still. Oracle's Cloud@Customer answers the third by leaving the ledger where it is. Salesforce answers the second by selling outcomes rather than seats. The scores on this page reward the firms whose customers face the highest cost of leaving, and that cost is disclosed in the backlog rather than in the marketing.
