Good day folks!

AI investors are funding growth but becoming less forgiving. Cerebras raised its outlook and still lost 16% after hours; Cisco and Nebius showed networking and power remain scarce; and coding startups attracted $543 million. Meta, meanwhile, entered a consequential youth-safety trial as its smart glasses drew a criminal complaint in Germany.

1. Cerebras learns that “good” is no longer good enough

What happened: AI chipmaker Cerebras reported second-quarter sales of $180.11 million, up 74% year over year but below the $194.23 million analyst estimate cited by Reuters. Its cloud business roughly quadrupled to $126 million, while hardware sales fell to $54.1 million. The company raised its 2026 adjusted revenue forecast to $880–$890 million, yet its shares dropped 16% in extended trading.

Why it matters: Cerebras is one of the clearest public tests of whether a specialist can take meaningful inference business from Nvidia. The results were hardly a disaster. The reaction suggests that richly valued AI companies now need to beat expectations, widen margins and look relaxed while doing both.

What to watch: Cerebras must scale manufacturing more than tenfold this year to support demand, including its $20 billion multiyear OpenAI agreement. Watch cloud margins, customer concentration and whether on-chip memory provides the supply advantage management claims. Source: Reuters.

2. Cisco and Nebius show where the AI bottlenecks moved

What happened: Cisco took $4 billion in fourth-quarter AI infrastructure orders from hyperscalers, bringing its fiscal 2026 total to $9.3 billion. It expects $7.5 billion in related revenue in fiscal 2027. AI cloud provider Nebius separately reported $582.3 million in quarterly revenue, raised its 2026 contracted-power target to 5 gigawatts and said it had more than $40 billion in customer commitments.

Why it matters: Yesterday’s infrastructure story was about GPUs. Today’s is about everything around them: switches, power, land and contracts. Cisco’s networking orders and Nebius’s capacity commitments indicate that the buildout is broadening, even as the economics become increasingly hardware-heavy.

What to watch: Cisco’s shares fell despite strong guidance, reflecting high expectations. Nebius spent about $5.7 billion in the quarter and says it could sell all planned 2027 capacity at current terms. That is a company assessment, not a guarantee. Sources: Cisco results, Reuters on Nebius.

3. AI coding attracts $543 million—and a quality-control thesis

What happened: Stockholm-based Lovable raised $400 million at a $13.3 billion valuation, double its December valuation. CodeRabbit raised $143 million at a $1.5 billion valuation. Lovable turns natural-language prompts into software; CodeRabbit reviews code, explains changes and looks for vulnerabilities and maintainability problems.

Why it matters: The two rounds fit together unusually well. Investors are backing both the tools that generate more software and the tools meant to check what gets generated. Apparently “move fast and have another robot inspect it” is now an asset class. For a wider market comparison, see our 2026 AI coding assistant guide.

What to watch: Valuations are private-market judgments, not proof of durable revenue. Watch retention, enterprise adoption and whether independent review tools catch defects that model-native checks miss. Sources: Reuters on Lovable and CodeRabbit.

4. Anthropic is reportedly exploring a $6 billion Decart deal

What happened: Anthropic is in early talks to acquire Nvidia-backed Decart AI, according to a Reuters source. Bloomberg first reported that a deal could be worth about $6 billion. Decart develops AI infrastructure and optimization technology alongside models for real-time video editing and simulated environments. Anthropic declined to comment; Decart had not responded.

Why it matters: An acquisition would point to two priorities: lowering inference costs and adding specialized multimodal talent before a possible Anthropic listing. It would also be a sizable shift from hiring researchers one by one to buying an integrated team and stack.

What to watch: This is a reported negotiation, not an agreed transaction. Price, structure and timing may change—or the talks may end. Source: Reuters.

5. Meta faces its biggest U.S. youth-safety trial yet

What happened: Jury selection began in Oakland for a federal trial involving claims from California, Colorado, Kentucky, New Jersey and 29 states. The states allege Meta designed Facebook and Instagram to keep children engaged, misrepresented safety and unlawfully used children’s data. Meta denies the allegations. Opening statements are scheduled for August 18, with Mark Zuckerberg and Adam Mosseri expected to testify.

Why it matters: The states seek more than damages. Requested remedies include age restrictions, removal of infinite scroll, limits on notifications and deletion of algorithms and AI models trained on children’s data. A nationwide order could reach deep into product design and data governance.

What to watch: Meta says potential damages could reach $1.4 trillion, but the attorneys general have not disclosed a demand at that level. The advisory jury’s findings will inform the judge, who can accept or disregard them. Source: Reuters.

6. Meta’s AI glasses meet Germany’s privacy line

What happened: German digital-rights group HateAid filed a criminal complaint against Meta, EssilorLuxottica units and four retailers over Ray-Ban Meta smart glasses. It argues that the devices violate a law barring sales of communication equipment designed for unnoticed recording. Frankfurt’s digital-crime unit confirmed receipt and will conduct a preliminary review.

Why it matters: Smart glasses turn privacy from a settings-page question into a bystander problem. Germany’s network regulator says the devices are not automatically banned when recording is clearly signaled, so the dispute may hinge on whether that signal is genuinely noticeable in ordinary use.

What to watch: A complaint is not a finding of wrongdoing. Watch whether prosecutors open a deeper investigation and whether regulators specify stronger recording indicators or sales rules. Source: Reuters.

The one thing to remember

AI’s next phase is being negotiated in three places at once: capital markets, physical infrastructure and courts. The technology can keep expanding rapidly while investors demand cleaner economics and governments demand clearer limits. Those forces are not contradictory; they are what an industry looks like when it stops being a science project.