500 Global

Building trust is a big challenge — and opportunity — in re-commerce. @Carousell's latest results show why. The company reported its first positive adjusted EBITDA, with revenue reaching US$141 million, up 18% year over year. We see this as validation of a long-term strategy. Over the past few years, Carousell has evolved beyond connecting buyers and sellers to building the trust layer that makes re-commerce work, with integrated payments, shipping, inspections, and buyer protection. AI has helped accelerate that shift. Today, sellers can create a listing with a photo, description, and suggested price in just three seconds, while AI also helps detect fraud and phishing to make transactions safer. Congratulations to Marcus Tan, @siurui, and the entire Carousell team on this milestone! Full story here: scmp.com/week-asia/life… @vishalharnal @khailee @AriefJohan @atxfrench

The future of AI will not be determined solely by who builds the smartest model." — Ren Ito, Co-founder & Chairman of @SakanaAILabs. In a new op-ed, Ren explores how recent geopolitical developments are reshaping the AI landscape. As access to frontier AI becomes increasingly strategic, the focus is shifting from building a single leading model to orchestrating multiple models into trusted, resilient AI systems. Trust, he argues, is no longer defined solely by model safety or performance. It also depends on reliable access, meaningful control over data, and governance that reduces dependence on any single AI provider. Read the full piece: japannews.yomiuri.co.jp/editorial/outs… @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee

We're excited to see @DeepInfra expand its global footprint with its first international data center in Toronto! The new 1.7 MW facility, equipped with more than 1,000 NVIDIA Blackwell B300 GPUs, is DeepInfra's ninth data center and will help enterprises run AI workloads with lower latency and greater efficiency by bringing compute closer to users and data. We look forward to seeing the team's continued investment in the infrastructure needed to support the next wave of production AI. Full story here: globenewswire.com/news-release/2… @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee @khailee @AriefJohan

As one of the world's oldest and premier AI research conferences, ICML brings together many of the researchers shaping the future of AI. That's why we were proud to co-host AI Super Night: ICML 2026 Edition alongside our portfolio companies @ElorianAI and humans&, bringing together first authors of accepted ICML 2026 papers alongside researchers from @nvidia, @AnthropicAI, @OpenAI, @Meta, @Apple, @GoogleDeepMind, and leading AI startups. Our lightning panel featured @AndrewDai (Elorian AI), @Niloofar Mireshghallah (humans&, @CarnegieMellon), @DZhang50 (@RicursiveAI), @albe_alfa (@axiommathai), and @meggmcnulty (Cosmic Labs), who shared their perspectives on where frontier AI research is heading. The conversations continued throughout the evening as researchers, founders, and builders connected in a closed-door setting. A big thank you to our collaborators Elorian AI, humans&, Cosmic Labs, @gmi_cloud, AttentionX, and @xiaohongshu for helping bring this community together. @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee

Congratulations to @SakanaAILabs Co-founder & Chairman Ren Ito on being appointed a Founding Member of the United Nations' AI for Good Global Commission. Joining leaders including Jensen Huang (@nvidia), @Benioff (@salesforce), @AlbertBourla (@pfizer), and Mukesh Ambani (Reliance Industries @RIL_Updates), Ren will help advance global discussions on AI trustworthiness, governance, and responsible social implementation. We're proud to see Sakana AI contributing not only to frontier AI research through its collective intelligence approach, but also to the international frameworks that will help ensure these technologies deliver broad economic and societal benefits. Read the full announcement: sakana.ai/ai-for-good-co… @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee @CourtneyPowell

From teams generating embeddings for knowledge bases to evaluating models and processing large datasets, many AI workloads don't require real-time responses. @DeepInfra's new Batch API is designed for these jobs, allowing developers to run asynchronous inference workloads at 20% lower cost than real-time pricing, with results returned within 24 hours. Because it's OpenAI-compatible, existing batch workflows can be reused with minimal changes. As AI moves further into production, matching infrastructure to the needs of the task is becoming an increasingly important way for teams to optimize performance and cost. Learn more: deepinfra.com/blog/batch-api @christine_tsai @TonyW @vishalharnal @atxfrench @khailee @AriefJohan

Great to see everyone at our @icmlconf event yesterday! If you missed it we have one more @ElorianAI event Thursday evening cohosted with @500GlobalVC and @xiaohongshu. Sign up here luma.com/vv164688

As AI moves from pilots to production, inference performance is increasingly measured by cost per token, not just raw compute. @DeepInfra is among the inference providers using NVIDIA's inference software stack to bring frontier open source models like DeepSeek V4 to Blackwell from day one. For enterprises and developers, that means immediate access to the latest open models on highly optimized infrastructure, reducing the time and effort needed to deploy production AI applications while benefiting from ongoing software improvements. @christine_tsai @TonyW @vishalharnal @khailee @AriefJohan

Happy 250th Birthday, America! 🇺🇸🎇🦅 As the child of immigrants, a second-generation American, sister of a U.S. Navy veteran, mom of two young citizens, and partner to so many brilliant founders building the future here, I am a fervent believer in the American Dream. May that dream continue to inspire for generations to come, from sea to shining sea.

50 years ago, my parents Michael and Jessie Wang immigrated to the U.S. 🇺🇸 from Taiwan. 🇹🇼 After several stints working as dishwashers and waiting tables, they saved enough to open a small coffee shop in the Tenderloin District of San Francisco called West Coast Coffee Shop. It was on 180 Golden Gate Avenue and Leavenworth. The rent they paid was $100 per month. I later learned there was a United Artists Studio office upstairs that gave them cheap rent so they could incentive my parents to open the coffee shop for their employees. My parents were the only ones working the restaurant, aside from my sister and I who occasionally helped after school. They built a loyal customer base over the years and the coffee shop was a bright spot in the SF Tenderloin. They eventually saved enough to open a Chinese restaurant in the Haight district called The Szechuan Court (pictured below). With the help of my aunts and uncles, my parents opened two more restaurants, each of which became iconic landmarks in their neighborhoods: The Metro in the Castro district (on 16th and Market) and Square nearby Union Square (Geary and Mason). On the 250th anniversary of the United States, I wanted to thank the U.S. for supporting entrepreneurs and small business owners like my family and giving us a chance to plant our roots here. Here’s to another 250 years of being the best startup country supporting people like my parents coming here and making a living. Happy 4th of July and Happy 250th anniversary to the United States of America! 🇺🇸

Japanese founders: if U.S. expansion is on your roadmap, but your GTM strategy still feels unclear, this program is worth a serious look. Applications close after the weekend: Monday, July 6 at 11:59 PM JST. The J-StarX Silicon Valley Extended Program is an 11-week, equity-free program for Japan-based startups with a working product, traction, and an established team. Selected founders will be supported with global market validation, fundraising preparation, pitch framing, U.S. expansion readiness, and access to global mentors, VCs, startups, and 500 Global’s and JETRO’s network. This is designed for founders who want structure, feedback, and a clearer path toward international growth. Final deadline: July 6, 11:59 PM (JST) Apply → be.500.co/japan-svep-b3 @JETRO_jgc @JETRO_info

Healthcare AI has a production problem, not an innovation problem." — @abhinavshashank, Co-founder & CEO of @innovaccer. It's a challenge worth solving: nearly 40% of every hospital dollar is spent on administrative work instead of patient care. The company's new multi-year Strategic Collaboration Agreement with @awscloud is designed to help health systems and payers bridge that gap by providing the infrastructure needed to deploy secure, healthcare-native agentic AI in production. The impact is already being seen. Community Care of North Carolina, which serves more than one million lives, was able to procure and deploy Innovaccer through AWS Marketplace in a fraction of the time required by a traditional procurement process, accelerating implementation and time to value. Full story here: businesswire.com/news/home/2026… @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee

Nearly US$1 trillion in announced sovereign compute commitments are now being made globally as governments treat AI infrastructure as a strategic asset. @meggmcnulty, Co-founder & CEO of Cosmic Labs, argues that sovereign compute isn't about building a domestic Nvidia or achieving full-stack AI independence. It's about retaining agency over critical layers of the AI stack while managing dependencies elsewhere. With frontier AI concentrated among a small number of labs, GPU clusters, and cloud providers, countries are increasingly viewing compute alongside energy, communications, and defense as foundational infrastructure. The real differentiator won't be capital commitments alone. It will be operational competence, utilization, and the ability to run production AI workloads effectively. Read the full piece: ciphertalk.substack.com/p/sovereign-co… @christine_tsai @TonyW @vishalharnal @atxfrench @leoclee

🔥 Congratulations @canva and team! Read more: robinhood.com/us/en/newsroom…

Entering the U.S. market with guesswork is expensive. For many Japan-based founders, the hard part is knowing which market signals matter, which customers to validate with, how to position the product, and how to explain the opportunity to a global VC audience. That is what the J-StarX Silicon Valley Extended Program is built for. Over 11 weeks, selected founders will work across Tokyo and Silicon Valley in a hybrid format to sharpen: 🔺 Global market validation 🔺 U.S. expansion readiness 🔺 Fundraising strategy 🔺 Pitch framing for international investors 🔺 Product feedback and global positioning 🔺 Access to mentors, experts, VCs, startups, and 500 Global’s network The program is non-investment and equity-free, designed for a curated group of Japanese founders ready to move closer to international growth. Best fit: Japan-based startups incorporated in Japan with a working product, traction, and an established team. Final deadline: July 6, 11:59 PM (JST) Apply → be.500.co/japan-svep-b3 @JETRO_jgc @JETRO_info

Today, Ambience is expanding its nursing suite with the introduction of Nursing Summary and Ambient Flowsheet Documentation. Together with Chart Chat for Nursing, released earlier this year, these three capabilities advance our single system of intelligence to support inpatient nurses throughout the entire shift.

Building a global startup from Japan has never been more possible. That doesn't mean it's easy. The J-StarX Silicon Valley Extended Program gives Japan-based founders 11 weeks to work on that next step: GTM, fundraising, outreach, and US market validation. It starts in Tokyo → Then 4 weeks in Silicon Valley with investors, customers, and operators who are there to engage with what you're building. You leave with a network and market validation that's hard to get any other way. 🔗 Apply now: be.500.co/japan-svep-b3 ⏳ Final deadline: July 6, 11:59pm (JST) @JETRO_jgc @JETRO_info

@Citadel @DeepInfra @nikolaborisof That’s why we were proud to co-lead @DeepInfra’s $107M Series B alongside @gharik w/ participation from @nvidia, @SamsungNext, @Supermicro & others. Together, they have built a small but mighty team on a mission to harness the power of AI and deliver intelligence to all of us.

@Citadel @DeepInfra DeepInfra recognized this shift early. As @nikolaborisof (Co-Founder) shared on Bloomberg, the company has spent the last four years building a purpose-built inference cloud optimized specifically for serving models in production.

While much attention has been given to training frontier models, recent developments suggest that the economics of deploying AI is just as important as the models themselves. Why we co-led @DeepInfra's $107M Series B alongside Georges Harik w/ participation from @nvidia 🧵