Cerebral oxygen delivery is central to the modern intensive care of patients with severe traumatic brain injury. Low brain tissue oxygen tension (PbtO2) results from microvascular collapse and diffusion limitation and is associated with adverse outcome. A number of therapies to improve oxygen delivery are known to be effective in improving PbtO2. Their relative effectiveness and microscopic regions of hypoxia, however, may exist/persist even in the presence of normal PbtO2. Unfortunately, there are no methods currently for assessing this quantitatively. We used an in silico (computational) simulation approach to understand the effect of common interventions on the microscopic distribution of brain tissue oxygen tension. We constructed a non-linear mathematical model of cerebral oxygen supply, diffusion, and consumption for a simplified geometry. Model parameters were chosen to agree with clinical parameters. We found that it was possible to create a plausible diffusion-limited scenario with a significant hypoxic fraction by increasing the mean diffusion distance. We found that increasing cerebral blood flow/blood oxygen content or suppressing the cerebral metabolic rate were most effective at improving PbtO2 and reduced the hypoxic fraction. Within the limitations of our modeling assumptions, increasing the arterial oxygen partial pressure was less effective and only improved PbtO2 by creating a region of hyperoxic tissue with no improvement in hypoxic fraction. The in silico simulations can be useful in understanding the likely physiological effect of complex treatments for which measurement techniques do not exist.