diff --git a/docs/src/guides/use_multithreading.md b/docs/src/guides/use_multithreading.md
index 61eccf775..cc5d61026 100644
--- a/docs/src/guides/use_multithreading.md
+++ b/docs/src/guides/use_multithreading.md
@@ -78,3 +78,113 @@ Even if you have a large number of nodes, using more threads than you have
 physical cores is likely to lead to bad performance. For example, on a quad-core
 laptop, using `--threads 52`, is likely to lead to bad performance, even if you
 have a graph with 52 nodes.
+
+## Threading with Gurobi
+
+Gurobi's `Gurobi.Env` objects are not thread-safe. This means that you cannot
+use the `optimizer = () -> Gurobi.Optimizer(env)` syntax if you are also using
+multithreading. (You can use `optimizer = Gurobi.Optimizer` because this creates
+a new environment for each node, but some license types have a limit on the
+number of concurrent licenses that can be created.)
+
+As a work-around, use the following function to partition the nodes into `N`
+disjoint sets and use a shared lock across the nodes within a group so that at
+most one subproblem within each group can be solved at any one time.
+```julia
+function shard_gurobi_licenses!(model::SDDP.PolicyGraph, groups...)
+    for group in groups
+        env, lock = Gurobi.Env(), ReentrantLock()
+        for node in group
+            set_optimizer(model[node].subproblem, () -> Gurobi.Optimizer(env))
+            set_silent(model[node].subproblem)
+            model[node].lock = lock
+        end
+    end
+    return
+end
+```
+For example, if you have two licenses you may want to group the nodes into "odd"
+and "even":
+```julia
+julia> shard_gurobi_licenses!(model, 1:2:52, 2:2:52)
+```
+or, if you have three licenses, you may want to split the nodes into "start",
+"middle", and "end" sets:
+```julia
+julia> shard_gurobi_licenses!(model, 1:17, 18:35, 36:52)
+```
+
+Here's a full example:
+```julia
+julia> using SDDP, Gurobi
+
+julia> function shard_gurobi_licenses!(model::SDDP.PolicyGraph, groups...)
+           for group in groups
+               env, lock = Gurobi.Env(), ReentrantLock()
+               for node in group
+                   set_optimizer(model[node].subproblem, () -> Gurobi.Optimizer(env))
+                   set_silent(model[node].subproblem)
+                   model[node].lock = lock
+               end
+           end
+           return
+       end
+shard_gurobi_licenses! (generic function with 1 method)
+
+julia> model = SDDP.LinearPolicyGraph(;
+           stages = 52,
+           lower_bound = 0.0,
+       ) do sp, t
+           @variable(sp, 0 <= x <= 100, SDDP.State, initial_value = 0)
+           @variable(sp, 0 <= u_production <= 200)
+           @variable(sp, u_overtime >= 0)
+           @constraint(sp, demand, x.in - x.out + u_production + u_overtime == 0)
+           Ω = [100.0, 300.0]
+           SDDP.parameterize(ω -> set_normalized_rhs(demand, ω), sp, Ω)
+           @stageobjective(sp, 100 * u_production + 300 * u_overtime + 50 * x.out)
+       end;
+
+julia> shard_gurobi_licenses!(model, 1:2:52, 2:2:52)
+
+julia> SDDP.train(model; parallel_scheme = SDDP.Threaded())
+-------------------------------------------------------------------
+         SDDP.jl (c) Oscar Dowson and contributors, 2017-26
+-------------------------------------------------------------------
+problem
+  nodes           : 52
+  state variables : 1
+  scenarios       : 4.50360e+15
+  existing cuts   : false
+options
+  solver          : Threaded()
+  risk measure    : SDDP.Expectation()
+  sampling scheme : SDDP.InSampleMonteCarlo
+subproblem structure
+  VariableRef                             : [5, 5]
+  AffExpr in MOI.EqualTo{Float64}         : [1, 1]
+  VariableRef in MOI.GreaterThan{Float64} : [4, 4]
+  VariableRef in MOI.LessThan{Float64}    : [2, 3]
+numerical stability report
+  matrix range     [1e+00, 1e+00]
+  objective range  [1e+00, 3e+02]
+  bounds range     [1e+02, 2e+02]
+  rhs range        [1e+02, 3e+02]
+-------------------------------------------------------------------
+ iteration    simulation      bound        time (s)     solves  pid
+-------------------------------------------------------------------
+         1   1.440000e+06  1.383984e+06  2.383709e-02       309   3
+        22   1.607500e+06  1.432500e+06  1.177620e+00     13832   2
+        62   1.405000e+06  1.432500e+06  2.469422e+00     30472   3
+-------------------------------------------------------------------
+status         : simulation_stopping
+total time (s) : 2.469422e+00
+total solves   : 30472
+best bound     :  1.432500e+06
+simulation ci  :  1.430250e+06 ± 3.113414e+04
+numeric issues : 0
+-------------------------------------------------------------------
+```
+
+!!! tip
+    If you need help with Gurobi licenses and multithreading, please open a
+    GitHub issue.