Alex Lisle\u2019s desk plants are in trouble\u2014they\u2019re next to four AI computers that have been running hot lately.<\/p>\n
Lisle, CTO at cybersecurity platform Reality Defender, has a quartet of DGX Spark machines, all featuring a Nvidia GB10 Grace Blackwell Superchip. Each workstation hosts a different large language model (LLM) to create what coders call an adversarial loop: models like Qwen3.8b write the code, and DeepSeek 4.1 Flash audits any flaws in the generators\u2019 work, with the aim of improving it.<\/p>\n
\u201cThey\u2019re all different models and they all hate each other, so it\u2019s effectively just a way of replacing the human to make good code for mundane tasks,\u201d he said, which means jobs like fixing workflows and reports.<\/p>\n
Lisle had tried the adversarial loop with cloud-based models and token-based pricing, which led to high costs as the LLMs went eight or nine rounds before concluding.<\/p>\n
\u201cThat was costing millions of tokens,\u201d he said. \u201cThe only reason that adversarial loops work for me is the fact that I can do this without paying for tokens.\u201d<\/p>\n
Lisle isn\u2019t the only tech pro considering the local train these days, experimenting with in-house hardware to support AI workflows. But this bring-your-own-AI-device option, which requires hosting open models, has IT pros juggling the cost benefits with security concerns.<\/p>\n
Keep it open? Tech companies like Google, Anthropic, and OpenAI offer large-scale models closed from download or modification. A paper from an MIT and Georgia Tech researcher, published in November 2025, found closed models account for 80% of model usage.<\/p>\n
Open-source models, however, are free and provide access to:<\/p>\n
Training data<\/a> and its sourceThe source code training and running the modelThe parameters supporting decision-making, or \u201cweights\u201d (\u201copen-weight\u201d AI does not necessarily mean open-source)<\/p>\n Open models running on one\u2019s own hardware eliminate unpredictable, per-token costs<\/a>. AT&T\u2019s chief data and AI officer told WSJ in August<\/a> that a quarter of the company\u2019s AI usage now comes from open models. (Other organizations are onboard, too, as IT Brew reported<\/a>.)<\/p>\n On the road. Parker Johnston, agentic field CTO at solutions integrator Insight, recently chronicled his road trip<\/a> picking up a brand-new ZGX Fury GB300 from HP\u2019s factory. He used the Fury hardware\u2014a station with a Nvidia Blackwell Ultra\u200b GPU and 748 GB of coherent memory\u2014to run four large language models<\/a> (LLMs) simultaneously and 30-plus agents; he burned 11 million tokens in 90 minutes, according to his LinkedIn post.<\/p>\n Johnston sees owning AI hardware as a way to help employees explore beyond the token-based constraints of cloud-based, third-party AI services.<\/p>\n Top insights for IT pros<\/p>\n From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.<\/p>\n \u201cFrom 6 pm to 6 am I have agents that can run now completely overnight and give my development team a workload or a capability they never thought possible,\u201d Johnston told IT Brew, citing tasks like vulnerability management<\/a> or spotting dependency drift, which is when a system uses different versions of the same package. \u201cAll of these things\u2014you might shy away from\u2026because one, it takes labor time to do it, and two, it takes tokens.\u201d<\/p>\n While Johnston notes the importance of putting any model behind a corporate network with access controls, he sees local models helping industries handling sensitive information.<\/p>\n \u201cMaybe it\u2019s a bank-loan application, and I can go onto this device, process the information I need, and immediately remove the files. I no longer have to worry about where they\u2019re being kept, what access folks might have, who configured the cloud environment, what\u2019s going on on the other side of it. It\u2019s all for me to now wholly own across that entire landscape,\u201d he said.<\/p>\n AI everywhere. Neil Ward-Dutton, research VP of agentic automation and AI technologies at tech-intelligence firm IDC, sees businesses running models on modest, local hardware options. Tools like Ollama help open-weight models run on a MacBook pro, Chrome includes an on-device LLM, and today\u2019s PCs have specialized GPUs for on-device GenAI tasks, he mentioned.<\/p>\n Ward-Dutton doesn\u2019t envision a future, however, where every employee has an expensive, chunky workstation running their AI: \u201cDo I see a future where enterprises will have AI models running in lots of different places, including on laptops, company phones, local private clouds, on-premise data centers<\/a>, and so on? Absolutely.\u201d<\/p>\n