MSDINTERNATIONAL← Home Discover MSD ⌄From AI and robotics to islands, yachts, specialist partners and new business models. Choose your entry point – and discover what lies behind it.
AI, robotics & futureFuture · Die Firma aus Systemen10,000 AI agents as a scaling target, networked tasks and an interactive estimator.
↗
AI, robotics & futureRobotics · 10 platformsManufacturers, capabilities, interfaces and integration tasks.
↗
MSD Direct & MarktplatzMSD Direct · Das PortalPeople, opportunities, way of life.
↗
MSD WORLDSiteWarden · Robotics LabGripping, sensor technology and physical AI – planned experimental programs.
↗Die Übersicht wird geladen.
MSD / BUSINESS / RESEARCH EDITION
Automation is a business decision: investment, usable performance, ongoing operations and real demand must fit together. A minute saved is not a profit. Only when this results in additional capacity, less effort or better customer results can the benefit be economically evaluated.
RECHERCHESTAND 12.11.2026 · FAKTEN + EINORDNUNG + SZENARIENMSD / BUSINESS / DARUM GEHT ES
The invoice does not just include the machine or model access. Setup, data preparation, interfaces, training, acceptance, energy, maintenance and failures change the profitability. On the benefits side, what counts are good items, completed processes and avoided rework - after deducting the human work that is still necessary.
If several providers use the same technology, customers can receive part of the savings through falling prices. Platforms, investors and employees have their own demands. Anyone who reliably solves a bottleneck or offers a special service can position themselves differently than someone who just sells the same routine at a lower price.
We start with a specific process and a reliable basis for comparison. What service is being provided today? Where do errors and waiting times arise? Which pilot can demonstrate an advantage? Only then are technical options and individual project costs compared. The website shows connections and model values, not flat-rate customer prices.
DATENFENSTER / KAPITAL, DAS BEREITS FLIESST
Four corporations reported a combined $357.508 billion in cash additions to property, plant and equipment in their fiscal years ending in 2025. In the respective periods of the previous year it was 217.267 billion. Own addition of the cash flow items mentioned below; nominal increase around 64.5%.
| Company | 2024 / FY2024 | 2025 / FY2025 | Operativer Cashflow 2025 / FY2025 |
|---|---|---|---|
| Microsoft | 44,477 | 64,551 | 136,162 |
| Alphabet | 52,535 | 91,447 | 164,713 |
| Meta | 37,256 | 69,691 | 115,800 |
| Amazon | 82,999 | 131,819 | 139,514 |
Delimitation: Tangible asset payments from the cash flow statement, not enterprise value, cash balance or exclusively AI budget. Microsoft: fiscal year to June; the others: calendar year. No completely identical period comparison group. Gross payments without deducting sales proceeds, without adding separately reported lease repayments and without non-cash asset additions. Meta's frequently cited 72.22 billion includes lease repayments; The 69.691 billion of the property, plant and equipment item is consistently shown here. The sum is an order of magnitude of the four selected reports, not a global AI investment amount.
The table shows the operating cash inflow next to it. It is not a freely available profit: taxes and parts of current business are already included, further investments, debt service and distributions compete for liquidity. Additional financing can come from inventory or debt and equity capital.
Historical acquisition payments do not predict the current sales value or the economic benefit of the investments. Depreciation spreads costs; Market value and company value depend on other variables. The net present value of an investment is calculated as discounted future net inflows minus initial expenditure.
Our conclusion: Not every company has to own infrastructure on a corporate scale. Controlled access, good data, appropriate integration and a measurable customer result will be interesting. Large capital flows show competition for capacity; they do not guarantee a return on investment for any project.
Microsoft · Fiscal year up to 30.06.2025 or 30.06.2024 · Item “Additions to property and equipment”. Kapitalflussrechnung im Original ↗
Alphabet · Kalenderjahr 2025 bzw. 2024 · Position „Purchases of property and equipment“. Kapitalflussrechnung im Original ↗
Meta · Kalenderjahr 2025 bzw. 2024 · Position „Purchases of property and equipment“. Kapitalflussrechnung im Original ↗
Amazon · Kalenderjahr 2025 bzw. 2024 · Position „Purchases of property and equipment“. Kapitalflussrechnung im Original ↗
Original values in millions of US dollars divided by 1,000. Addition and rate of change calculated editorially. Historical nominal values; no conversion to today's purchasing power and no exchange rate calculation. The interactive model below explains purchasing power under freely selectable assumptions.
01 / THE ECONOMICAL MACHINE
Own funds or credit enable the development. Credit also creates an obligation to repay.
Hardware, software, data, training and integration initially tie up resources.
More usable output per input – after scrap, rework and downtime.
Competition, contracts and bargaining power distribute the savings.
Only realized cash flow pays for repayments, operations and the next investment.
Geldgrundlage: ECB · Money and credit ↗ · Organisationsinvestitionen: The productivity J-curve ↗
02 / DIE LETZTEN JAHRZEHNTE
The open web lowers barriers to distribution. Infrastructure and access become prerequisites for new products. Source: CERN, web release 1993.
Digital products combine high development costs with often low additional distribution costs. Network effects can promote concentration; this is a mechanism, not a law for every platform.
Language models and agents access tools. Costs per task can decrease; Audit and context become more valuable. At the same time, the study shows large differences between tasks.
MSD scenario: Data, power, connected load, secure integration and reliable service gain in importance. Whoever solves these bottlenecks could extract more value. The timing and winner remain open.
Web source ↗ · Productivity findings ↗ · IEA-Infrastrukturbericht ↗
03 / DIE VERTEILUNGSFRAGE
| Per piece | Vorher | Vorsprung | Preiskampf |
|---|---|---|---|
| Sales proceeds | 100 | 100 | 70 |
| Total unit costs | 80 | 60 | 60 |
| Covering surplus* | 20 | 40 | 10 |
* Free set, simplified model. “Surplus” here is revenue minus assumed total unit costs, not annual taxable profit. Assuming the same number of pieces. The cost reduction from 80 to 60 may be accompanied by a declining surplus under price pressure from 100 to 70.
5,179 Support staff form the basis of the NBER working paper. With assistance, the number of issues resolved per hour increased by an average of 14%. New and less experienced forces particularly benefited.
Brynjolfsson, Li & Raymond · NBER 31161Brynjolfsson, Rock, and Syverson explain how complementary investments in organization and knowledge can delay the measured benefits of an enabling technology.
NBER · Productivity J-CurveAcemoglu's task-based model connects affected activity shares with savings and overall economic productivity. The May version 2024 estimates at most 0.66% additional TFP over ten years.
Acemoglu · Modellfassung Mai 202404 / MONEY OVER TIME
Three different sizes: cash without interest, nominal income and real purchasing power. Technology influences the supply side; Interest rates, demand, market power and distribution have a say. A cheaper calculation does not automatically make housing or energy cheaper.
* Illustrative scenario. Starting index 100 in each case. No forecast of inflation, wages or wealth. Purchasing power = 100 / (1 + inflation)^years; nominal income = 100 × (1 + income growth)^years; real income = nominal income / (1 + inflation)^years. No interest, taxes or changes in the consumer basket. Two percent is a model default; the ECB has a medium-term target, not an annual guarantee.
EZB-Strategie 2025 ↗ · Check project costs separately in the future model* ↗
05 / NO EXCUSE FOR STOP
Don't just buy access to AI: think about data rights, export, termination, maintenance and failure scenarios.
Process knowledge, quality control, safety, interfaces and maintenance can gain in value. This does not replace concrete qualification planning.
Ask for reliable pilot data, capital commitment and recurring customer benefits. Technical demonstration and sustainable cash flow are different proofs.