Tomasz Tunguz

Student of Startups
Backer of 9 unicorns
Author of http://tomtunguz.com 
Subscribe https://ttunguz.kit.com/ef0a779b9c

Founder

I burned 84 million tokens on February 28th. Researching companies, drafting memos, running agents.

That’s running Kimi K2.5, a serverless model via API. At Claude or OpenAI rates — roughly $9 per million tokens blended — equivalent usage would cost $756 for a single day’s work. My peak days hit 80 million tokens. My average days run 20 million. Cloud inference at frontier-model pricing adds up fast.

This week, Alibaba released Qwen3.5-9B, an open-source model that matches Claude Opus 4.1 from December 2025. It runs locally on 12GB of RAM. Three months ago, this capability required a data center. Now it requires a power outlet.

A $5,000 laptop — a MacBook Pro with enough memory to run Qwen locally — pays for itself after 556 million tokens. At my usage rate, that’s about a month. At 20 million tokens per day, it’s four weeks.

After payback, the marginal cost drops to electricity.

It isn’t an intelligence compromise. Reasoning, coding, agentic workflows, document processing, instruction following : the 9B model matches December’s frontier across the board.

What changes when frontier intelligence runs locally? Everything I send to cloud APIs today — drafting emails, researching companies, writing code, analyzing documents — stays on my machine. No API logs. No third-party retention. No outages. No rate limits.

The tradeoff is parallelization. Cloud APIs handle thousands of concurrent requests. A laptop runs one inference at a time. For simple tasks — summarization, drafting, Q&A — that’s fine.

Queue them up. Let them run overnight. For complex agentic workflows that spawn dozens of parallel threads, local inference may not be worth the wait. The economics favor depth over breadth : fewer tasks, run longer, run cheaper.

Three months from data center to laptop. The buy-vs-rent math just changed.

Databricks started later. It built a more complex architecture. It focused on unstructured data; images, documents, logs, audio. Though vast within the enterprise, this data had historically produced little insight. Too hard to process. Too messy to query. Too expensive to store in formats that mattered.

Snowflake took the opposite bet. Structured data. Clean tables. SQL queries that ran fast & returned answers executives could read. The market agreed. Snowflake went public at a $70 billion valuation. Databricks raised private rounds at half that.

Then AI arrived. Suddenly the data that was too messy to query became the data that models needed to train. Unstructured data wasn’t a liability. It was the asset.

Databricks started later. It built a more complex architecture. It focused on unstructured data; images, documents, logs, audio. Though vast within th

Open article: Databricks Overtakes Snowflake by Tomasz Tunguz • 2 min read

Last quarter, my AI inference costs hit $100,000 annualized.

I started small. Six months earlier, I was spending $200 a month on Claude. Then I added three agent subscriptions : Codex, Gemini, & Claude Code. I was paying $600 a month.

Next I started using AI to transform my todo list into my done list, increasing tasks to 31 per day. $92 daily inference invoices started arriving. Then $400 per month on browser agents.

Within two quarters, my inference spend grew from $7,200 to $43,000 to over $100,000 run rate.

Last quarter, my AI inference costs hit $100,000 annualized. I started small. Six months earlier, I was spending $200 a month on Claude. Then I added 

Open article: Will I Be Paid in Tokens? by Tomasz Tunguz • 2 min read

I burned 84 million tokens on February 28th. Researching companies, drafting memos, running agents.

That’s running Kimi K2.5, a serverless model via API. At Claude or OpenAI rates — roughly $9 per million tokens blended — equivalent usage would cost $756 for a single day’s work. My peak days hit 80 million tokens. My average days run 20 million. Cloud inference at frontier-model pricing adds up fast.

I burned 84 million tokens on February 28th. Researching companies, drafting memos, running agents. That’s running Kimi K2.5, a serverless model via A

Open article: Data Center Intelligence at the Price of a Laptop by Tomasz Tunguz • 2 min read

Data Center Intelligence at the Price of a Laptop

For the first time in venture history, three distinct channels share the liquidity burden roughly equally.

A decade ago, secondaries barely registered. They accounted for roughly 3% of exit value in 2015. Today they claim 31% : nearly $95b in the trailing twelve months.

The shift accelerated after 2021’s IPO bonanza. When public markets closed their doors in 2022, investors found alternative routes. Secondaries absorbed demand that would have flowed to traditional exits. When Goldman Sachs acquired Industry Ventures, the transaction signaled secondaries have arrived. Morgan Stanley followed with EquityZen, then Charles Schwab announced its acquisition of Forge Global. Wall Street recognized the structural change before most of venture did.

This matters for founders & investors. When IPOs dominated exits, fund models assumed a small number of public offerings would generate the bulk of returns.

Now liquidity arrives through multiple doors. A founder might sell secondary shares to patient capital while the company remains private. A GP might move positions through continuation vehicles. An LP might trade fund stakes on an increasingly liquid secondary market.

The 830 unicorns holding $3.9t in aggregate post-money valuation cannot all exit through IPOs. The math doesn’t work. At 2025’s pace of 48 VC-backed IPOs, clearing the unicorn backlog would take seventeen years. Secondaries provide a release valve that traditional exits cannot.

Companies like OpenAI have embraced this reality, running employee tender offers while voiding unauthorized secondary transfers. The largest private companies now manage their own liquidity programs rather than waiting for public markets.

Today, secondary liquidity concentrates in the top 20 names. SpaceX, Stripe, OpenAI. For the founder of company #50, the secondary market remains largely theoretical. For secondaries to succeed as a broad asset class, buyers must underwrite positions in companies without household recognition. As the market grows, this coverage gap becomes opportunity.

For LPs starved of distributions since 2022, the expansion of secondary channels offers hope. The $169b in cumulative negative net cash flows needs somewhere to go. More exit paths mean more opportunities to return capital.

When a Series B employee asks about liquidity today, the answer isn’t “wait for the IPO.” It’s “we’re planning a tender offer next year.”

A decade ago, secondaries were a footnote. Now they’re infrastructure. Liquidity flows where it can, not where tradition suggests it should.

Stealth Mode Startup Company

NodePrime

Responsible for product design, requirements and build budget. Using agile methodologies to build an hyperscale SOA product.

Agile Project Management
Product Architecture, Design and Engineering.
Internships: Program design, Talent Engagement and Talent Management.
Talent acquisition.
Patent Drafting
Gap analysis
Hire Planning
Document Driven Development

Vice President of Engineering

A start-up building and developing a SPoG bare-metal to Application layer (OSI full stack) infrastructure management and big-data analytics with machine learning solution. 

Agile Project and Product Management.
Product Design and Engineering.
Internships:  Program design, Talent Engagement and Talent Management.
Talent acquisition.
SBIR Program Application Drafting.
VC Engagement.
Financial Modeling.
Business Model Development.
Go-to-market strategy.
Partner Engagement.
Patent Development and Applications.
Team Management.
Market analysis.
Strategic Market Positioning.
Gap Analysis

Vice President of Business Development

VMware

Architect Infrastructure needed to support cloud-scale customer  solutions testing. Performing resource gap analysis and researching hardware requirements and capabilities for compute, networking and storage. Working with cross-functional teams and negotiating with vendors.

Snr MTS - Cloud Infrastructure Architect and Engineer

Automate the deployment and management of multiple virtual infrastructure test beds for VCOPS 5.6 at cloud scale (50K+ VMs each). 

VCOPS is a cloud operations management and monitoring tool.

Snr MTS - Lead Software Engineer: VCOPS 5.6 Virtual Infrastructure Engineer

Managed team to automate cloud scale ((30K+ VMs) test beds for VCOPS 5.0.x releases.  

Developed infrastructure tracking and remediation work-flows. 
Developed tools to mange and maintain cloud test beds controlling  250K VMs on 240 physical hosts connected to 150 TB of hybrid SAN protocol storage.

Snr MTS - Software Engineering Lead: VCOPS 5.0.x Virtual Infrastructure Engineer

This was an intern project to explore examine to technical challenges of providing a self-service cloud scale infrastructure deployment tool. Involved mentoring the intern and guiding the him through technical and design challenges.

Snr MTS - Software Engineering Project Manager: SIaaS

Added configurable and extensible full metric set and value shaping simulation for the virtual infrastructure simulator.

Snr MTS - Lead Software Engineer: Virtual Infrastructure Simulator Extensions

Developed Cloud Infrastructure management automation.
Developed UI testing automation using the Selenium action based frame work.
Created DOEs for software engineering  performance and scale characterization.
Created variance analysis reports.
Completed VCP 4.0 training

Snr MTS -  Lead Software Engineer: CIQ 1.x Virtual Infrastructure Engineer

Stepped in to cover for a team member's family emergency. Raised and resolved several issues with the product delivery process. The team ultimately delivered infrastructure management and test automation using STAF/STAX and Java as well as scalability and performance metrics and reports.

Lead Test Automation Engineer

Gathered Requirements across organization.
Performed gap analysis of team skill-set vs software engineering  requirements.
Found and hired talent into the team.
Technologies: GWT, Spring, Java and Hibernate
Developed Prototype.

Snr MTS - Lead Software Engineer: TCMS Program Lead

Brought in to bring out-of-control project back on track. Designed and implemented extensions to FOSSology tool. 
Worked across functional teams; product and release management and legal, to gather requirements.
Implemented training and support programs for product and release managers.
Brought project back on schedule - saved company ~ $500K in vendor fees.

Snr MTS - Software Engineering Lead: FOSSology

Magzor Corporation

Providing product, patent portfolio and business advice: 
Refocused the business on non-invasive medical devices.
Provide technology assessment. 
Introduced  and implemented agile team and project management.

Business Adviser - VP Software Engineering

Analatom Inc

Analatom provides research services to government and commercial agencies in the area of Structural Health monitoring, 

I have been helping to transition from a purely research services based business to developing commercial products and solutions from this work. I provided the vision and architectural design for the Asset Cycle Engineering (ACE) system, which is the flagship product for the company. 

Managed delivery of a prototype real time asset management and tracking system integrated with mobile device (smart phones and tablets). 

Assisted with writing research proposals (SBIRs) as well as technical support for systems integrations and development.

Business Adviser  - Vice President of Engineering

Ektasis

Developed Cloud based auction site aggregation tool. A High performance, cloud scale product that cached and compared auctions from different sites and allowed analysis of historical pricing info and identified auctions parameter that gave arbitrage opportunities and could automatically snipe (bid in the last second) on those auctions.

Developed a JIT high performance multi-threaded cache that stored over 90K references to a single disk in less than  a minute.

Developer: Lead Software Engineer for Bid Beagle

Modified Ebay Browser and MS Windows OS toolbar binaries and injected cloud application deployment functionality for our cloud based application management and launch product.

Developer: Lead Software Engineer: Cloud App Deployment from Ebay Toolbar

National Semiconductor

Architect for Development of an Enterprise Solution for end-to-end Test Chip Development and Automation. The tool brought down process development times by allowing engineers to automatically generate process control structure layouts, automated electronic test code and documentation.  

During this work I also Invented novel devices in the fields of Spintronics and semiconductor processing.

Principal Device Engineer: Principal Software Engineering for Automation

Principal Engineer responsible for the invention of MEMS/NEMS devices and processes.  Worked  with University Alliance programs and managed teams at IIT and OSU.  Numerous patents and technologies were developed.

Principal Device Engineer: Principal  MEMS/NEMS device and process engineer

Experienced Founder with  knowledge of information technology and services industry. Expertise in Patents, Software as a Service (SaaS), Scrum, Big Data, ML, and Recruiting. Entrepreneurial  professional with a PhD in Physics 

taplio

The Battle

Tomasz Tunguz

Mike Mian

Followers

Posts

Engagements

Likes

Performance ScoreTomasz Tunguz's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Tomasz Tunguz

Mike Mian

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Tomasz Tunguz's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreTomasz Tunguz's LinkedIn Performance