Writing

Evidence from a survey of 19 organizations with production AI deployments on which providers they use, how hard it would be to switch, and what keeps them locked in.

The second part of a series on Switzerland's new SIEC test explains, using oligopoly theory, how a merger can raise prices without creating a dominant firm. It introduces concentration measures, diversion ratios, and upward pricing pressure as tools for assessing these effects, which still have to be weighed against efficiencies.

Switzerland's revised Cartel Act replaces the dominance test in merger control with the EU's SIEC test. The post argues that this modernizes the method rather than tightening the law: it closes gap cases, considers efficiencies explicitly, and shifts the focus to a merger's effects on prices, quality, and innovation.

Revenue-management systems forecast demand and optimize prices but largely ignore how competitors will respond. The post argues that game theory, and compositional game theory in particular, can model these strategic interactions, and describes a pilot of competitor-aware dynamic pricing.

Most Swiss households and small firms pay fixed annual tariffs, so wholesale price swings never reach them. Drawing on a study for SECO, the article shows that end customers do respond to prices, with elasticities of about −0.2 for households and −0.4 for firms, and proposes tariffs that pass on more of the price signal while keeping certainty.

Training foundation models takes data, computing power, and expertise that few firms have, and those firms are vertically integrated across the AI value chain. The post argues that access to data may become an essential facility that competition law and regulation, including in Switzerland, will need to address.

Building competitive foundation models takes data and computing power that few firms control, and plug-in ecosystems add network effects that can entrench them. The article contrasts the EU's response (DMA, DSA, Data Act, AI Act) with Switzerland's lack of comparable rules and recommends that Swiss policy address access to data.

Reads the draft EU AI Act through a competition lens: its links to the DMA's gatekeeper concept and the separate obligations for foundation models, which do not fit the Act's risk-based scheme. The post notes that the EU is pursuing competition goals through regulation and asks how far it will go in intervening in markets.

Sorts the competition risks of AI into three settings: interaction between competitors (algorithmic collusion and herding), third-party arrangements such as hub-and-spoke, and vertically integrated intermediaries that favour their own offers. The post recommends that firms cover how their algorithms behave in the market within their responsible-AI frameworks.

Self-learning pricing algorithms can learn to avoid price wars and sustain high prices without communicating, a form of tacit collusion that competition law, which hinges on agreement or intent, may not catch. The article argues for closing this gap, for example by treating reward-punishment pricing strategies as collusion or by letting authorities test pricing algorithms before they are deployed.

Wholesale electricity demand barely responds to prices, but a 10% rise in end-customer prices cuts household demand by 1–3% and business demand by 2–7%. A survey of business customers finds that most passed higher electricity costs on to their own customers while working to cut consumption and invest in their own generation.

When hydropower concessions expire, the plants revert to the cantons and municipalities that granted them. The report reviews how the mountain cantons have approached these reversions, as background for the Canton of Zurich.

Quantifies the revenue potential of different types of hydropower plants and the costs of marketing their output, as input to the SFOE's work on a sliding market premium for hydropower.

Assesses the expected economic effects of introducing legal protection for journalistic content online, under which large online services would remunerate media companies for using their content.

Develops a terminology and framework for analysing data and digital policy challenges, with recommendations for Switzerland.

Reviews the measures European countries introduced against the 2022 energy-price spike and assesses which of them would suit Switzerland, as input to the Federal Council's decisions in December 2022.