AI Sector Daily Digest — July 14, 2026

1. PixVerse brings its Series C total to $439 million

• PixVerse, a Singapore-based AI video generation platform, said its Series C extension brings total Series C fundraising to $439 million. 1
• The extension added investors including Alibaba, Lollapalooza Capital, Ivy Capital, Grand Mount Capital, Eastern Bell Capital, Mirae Asset, BlueFocus, and CloudAlpha. 1
• PixVerse says it has more than 150 million users across 177 countries and is expanding its R1 real-time world model into interactive games and livestreamed AI characters. 1

2. LTM says AI-native revenue is already 12% of sales

• Indian software services firm LTM disclosed AI revenue for the first time: $150 million on a quarterly run-rate basis, or 12% of total revenue, across three AI-native businesses. 2
• CEO Venu Lambu told Reuters that LTM expects AI revenue to outpace traditional services as enterprises need help deploying large language models from firms such as Anthropic and OpenAI. 2
• LTM signed a Claude deployment partnership with Anthropic on Monday and says client concern has shifted toward controlling token costs and governance. 2

3. White House plans a broader AI power-cost pledge

• The White House plans to bring utility companies and data center developers into a voluntary pledge meant to keep AI-driven electricity demand from raising power bills for households and businesses. 3
• The planned event would broaden an earlier Ratepayer Protection Pledge signed by Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI. 3
• The policy problem is direct: state regulators and consumer advocates worry that local ratepayers could subsidize grid upgrades built mainly for large AI data centers. 3

4. Nvidia tightens Asia AI-chip buyer checks, FT reports

• Nvidia has more than halved the number of Asian customers authorized to buy its AI chips after introducing a tougher compliance white list, the Financial Times reported. 4
• The reported checks focus on Singapore, Malaysia, and Japan, with neo-cloud providers especially likely to have failed the first review. 4
• Reuters could not independently verify the report, but the move fits Washington's May guidance aimed at stopping advanced AI chips from reaching Chinese entities through third countries. 4

5. Hourglass reasoning isolates a planning bottleneck

• A new arXiv paper from Huan Zhu at Peking University introduces Hourglass reasoning, a workflow that forces LLMs to pass only a compressed symbolic state between reasoning stages. 5
• On ARC-AGI-2, Hourglass raised best-of-5 accuracy by up to 14 percentage points over an iterative self-refinement baseline; on ChipBench, it nearly doubled GPT-5.5 Verilog synthesis accuracy from 31% to 58%. 5
• The practical claim is that agents may need stricter context isolation, not just longer reasoning traces, when they must infer rules from a small number of examples. 5